Dementia knowledge and its demographic correlates amongst informal dementia caregivers in Singapore

Abstract

Objectives: The Dementia Knowledge Assessment Scale (DKAS) is a validated assessment tool to measure dementia knowledge. However, the factor structure of the DKAS has yet to be validated in Singapore. This study aims to investigate the DKAS factor structure amongst a sample of informal dementia caregivers in Singapore, as well as their sociodemographic correlates.

Methods: A total of 282 participants were evaluated on their knowledge of dementia by an interviewer administering the DKAS. Confirmatory factor analysis (CFA) of factor models proposed by previous study yielded poor fit for our sample. Thus, an exploratory factor analysis (EFA) was conducted. Multiple linear regression was then performed to examine the sociodemographic correlates of DKAS factors.

Results: EFA revealed a 23-item 3-factor model – ‘misconceptions about dementia’ (MD), ‘caregiving considerations towards dementia’ (CD) and ‘dementia symptoms’ (DS). Being a male caregiver and having lower educational levels were associated with poorer scores on MD and CD.

Conclusion: A 3-factor model of the DKAS was found to be more appropriate with the sample in this study. Findings from this study suggests the need for more dementia literacy intervention for caregivers of dementia patients in Singapore, especially for male caregivers and those of lower education levels.

Keywords:

• Dementia and cognitive disorders
• mental health
• dementia knowledge
• Dementia Knowledge Assessment Scale

Introduction

Dementia is a major global health concern of the twenty-first century (WHO, 2015) with no known cure currently for those afflicted with common forms of the disease (NHS, 2018). Taking into account that old age is one of the biggest risk factors for dementia (WHO, 2015), the prevalence of dementia is projected to increase substantially to as high as 115.4 million people (Prince et al., 2013), given the rise in populations’ median age globally (Lutz, Sanderson, & Scherbov, 2008). The ramifications of such a trend includes a significant increase in the global monetary costs associated with the disease (Wimo et al., 2017), as well as an increase in the number of people having to take on the role of an informal caregiver to a person with dementia, a role which had been reported to be a highly stressful (Tremont, 2011). These ramifications are also particularly relevant to Singapore, a country with an aging population and a 10% prevalence of dementia among those aged 60 years and above (Subramaniam et al., 2015). Due to the debilitating symptoms of dementia and the progressive nature of the disease, persons with dementia typically require considerable care as their condition deteriorates (Agüero-Torres et al., 1998). In this context, it has been reported that knowledge of dementia is important in facilitating the care management of persons with dementia (Chodosh et al., 2007; Gibbons, Teri, Logsdon, & McCurry, 2005; Matsuda et al., 2018).

Being knowledgeable about dementia entails several benefits for the care management of dementia patients. For instance, being more well-informed about dementia arguably enables one to better recognize the disease’s symptoms. Early detection and treatment of the disease provides several benefits such as slowing down patient’s cognitive decline (Gauthier, 2005; Rodakowski, Saghafi, Butters, & Skidmore, 2015), helping family or informal caregivers to better transit into the role of caregiving with early interventions (de Vugt & Verhey, 2013), thereby possibly improving the overall well-being of caregivers (Chiu, Wesson, & Sadavoy, 2013; Signe & Elmståhl, 2008). In addition, it has been reported that having higher levels of knowledge pertaining to dementia care positively influenced the quality of care provided to patients (Chodosh et al., 2007; Gibbons et al., 2005; Matsuda et al., 2018). A study by Graham, Ballard, and Sham (1997) demonstrated that while higher caregiver’s level of knowledge on dementia correlates to higher level of anxiety, it is also positively associated with their feelings of competency, reduction of expectations and increased use of positive comparisons, which consequently results in lower levels of depression (Graham, Ballard, & Sham, 1997). Conversely, a lack of knowledge in dementia conveys implications such as delay in diagnosis due to an inability to discern signs of normal aging from dementia symptoms (Rimmer, Wojciechowska, Stave, Sganga, & O’Connell, 2005), and higher subjective distress in caregivers due to overestimation of patient’s abilities (Sörensen & Conwell, 2011).

According to a meta-analysis, the aggregate prevalence of depression occurring amongst caregivers of dementia patients is 34% (Sallim, Sayampanathan, Cuttilan, & Chun-Man Ho, 2015). Depression also occurs more frequently amongst caregivers of persons with dementia as compared to those who care for patients with other chronic illnesses (Schoenmakers, Buntinx, & De Lepeleire, 2010). In this regard, it has been suggested that psychoeducational interventions which impart knowledge on coping with dementia symptoms may ameliorate caregiver burden and depression (Chiu et al., 2013; Gibbons et al., 2005; Ostwald, Hepburn, Caron, Burns, & Mantell, 1999). As such, scales that measures dementia knowledge would be useful in assessing an individual’s degree of knowledge in dementia, establishing baseline understanding of dementia, and determining the efficacy of educational programmes and interventions.

A review of five dementia knowledge measures underlined the limitations with all the five measures that were identified (Spector, Orrell, Schepers, & Shanahan, 2012). The review highlighted several issues such as the scales being outdated, lacking generalizability and possessing inadequate psychometric vigour. To address and overcome the limitations of the various dementia knowledge scales mentioned in the review, the Dementia Knowledge Assessment Scale (DKAS) was recently developed (Annear et al., 2015). The DKAS is a 27-item scale which has been established to be valid and reliable amongst a cohort that consisted of Australian medical students, members of Australians Health workforce and volunteers who participated in a dementia-related massive open online course (MOOC) (King, Robinson, & Vickers, 2014). In addition, a comparative study between DKAS and Alzheimer’s Disease Knowledge Scale (Carpenter, Balsis, Otilingam, Hanson, & Gatz, 2009) found that DKAS displayed superior internal consistency and higher sensitivity (Annear et al., 2016).

A confirmatory factor analysis (CFA) amongst an international cohort of volunteer respondents (from nine different countries) who had participated in a dementia-related online course indicated a 4-factor structure for the DKAS, with the number of items being reduced to 25 (Annear, Toye, et al., 2017). To our knowledge, with the exception of Japan (Annear, Otani, & Li, 2017), the DKAS has not been validated in any other country outside of Australia. However, a principal component analysis revealed that the Japanese translated version did not support the proposed 25 items 4-factor structure, but instead was more appropriate when used as an 18 items unidimensional scale (Annear, Otani, et al., 2017). Hence, this implies that despite the 25-item 4-factor structure of DKAS being validated on a cohort of international volunteers, the study might not have been sensitive enough to capture specific cultural and sociodemographic characteristics. By that same token, previous studies found that level of dementia knowledge can vary according to cultural and sociodemographic differences (Ayalon, 2013; Gray, Jimenez, Cucciare, Tong, & Gallagher-Thompson, 2009).

In view of these findings, our study aims to (1) examine the appropriateness of existing DKAS factor structure amongst a sample of informal caregivers of dementia patients in Singapore; and (2) explore the sociodemographic correlates of each DKAS factor and identify how these characteristics affect knowledge and level of understanding towards dementia amongst this sample.

Methods

Participants and procedures

A total of 282 participants were recruited over the period of January 2017 to December 2018, using a convenience sampling strategy. Participants recruited were primary informal caregivers of dementia patients, who were most involved in the care management of the dementia patient. Participants were recruited from two sites, namely the outpatient clinic of Institute of Mental Health (tertiary mental health provider) and its satellite clinic and a geriatric clinic in Changi General Hospital (tertiary general hospital). Inclusion criteria included: (1) being a Singaporean or Permanent Resident of Singapore who is at least 21 years of age and able to provide consent; (2) being the primary informal caregiver to a patient who has been formally diagnosed with dementia; (3) able to read or speak either English, Chinese or Malay language. After providing written informed consent, participants were administered a series of questionnaires designed to collect their sociodemographic information and to assess their dementia knowledge. The study was approved by the institutional ethics committee, the Domain Specific Review Board of National Healthcare Group in Singapore.

Translation of questionnaires

Although English is predominantly the official language in Singapore, many among the older adults do communicate more frequently in their second language and are therefore more fluent in their native tongue, especially for the Chinese and Malays. Therefore, the 27-item DKAS was translated into both Mandarin and Bahasa Melayu (official written language of Malays) following a standard ‘translation back-translation’ procedure (Beaton, Bombardier, Guillemin, & Ferraz, 2000). This was done to enable participants to be interviewed in their preferred language.

The translation process included having qualified individuals first do a single forward translation of the scale, which is then reviewed by a panel consisting of professionals and researchers. During this reviewing process, panel members flagged out any inadequate expressions in the translation, and identified discrepancies between translated and original version. The panel then discussed with study team members how these issues could be resolved. After which, back translation was also done to minimise the discrepancies. Pre-testing was conducted with some potential participants who indicated their preference for the survey to be conducted in the translated language over English. Feedback was gathered from the testers such as words/phrases in the translated scale that they felt were unclear, irrelevant or perhaps not comprehensible in the local context. The necessary edits were made to achieve better conceptual equivalency, and the final version was thereby developed.

Study questionnaires

The DKAS comprises 27 items which are statements pertaining to dementia with some being factual. All 27 items offer the same five response options: ‘false’, ‘probably false’, ‘probably true’, ‘true’ and ‘I don’t know’. Scoring for the 27 items works as such: for each item containing factual statements (e.g. ‘Exercise is generally beneficial for people experiencing dementia’), participants are accorded 2 or 1 point, if they responded with ‘true’ or ‘probably true’, respectively, while all other responses result in a 0 score. On the other hand, for items containing non-factual statements (e.g. ‘It is impossible to communicate with a person who has advanced dementia’), participants will be accorded 2 or 1 point, if they responded with ‘false’ or ‘probable false’, respectively, while all other responses are accorded 0.

A sociodemographic questionnaire comprising participant’s age, gender, marital status, relationship to care recipient and duration of caregiving was used to gather relevant data from the caregivers.

Both the DKAS and sociodemographic questionnaires were administered to participants during a face-to-face interview conducted by trained interviewers. This method of data collection was employed so that even participants with limited literacy would still be able to participate in the study with minimal potential of misinterpreting the DKAS’s items.

Statistical analysis

CFA were performed using model structures proposed by previous studies (Annear, Otani, et al., 2017; Annear, Toye, et al., 2017; Annear et al., 2015). A model is considered good if (1) the comparative fit index (CFI) >0.95; (2) the Tucker–Lewis index (TLI) >0.95 and (3) the root mean square error of approximation (RMSEA) <0.06 (Hu & Bentler, 1999). However, the model fit indices suggested that all three models fitted poorly with our sample (Hu & Bentler, 1997). Please refer to Table 1 for the model fit indices. Therefore, an exploratory factor analysis (EFA) was conducted in an attempt to construct a more appropriate factor structure for our sample.

Table 1. Model fits for various CFA structure of DKAS.

	Model 1	Model 2	Model 3
Chi-square	613.42	507.56	307.01
Df	318	269	135
CFI	0.82	0.83	0.81
TLI	0.80	0.81	0.79
RMSEA	0.057	0.056	0.067

Model 1, 27-item 4-factor structure (Annear et al., 2015); Model 2, 25-item 4-factor structure model (Annear, Toye, et al., 2017); Model 3, 18-item 1-factor structure (Annear, Otani, et al., 2017).

In the EFA, items with loading score of <0.3 were eliminated from the model. Appropriate number of factors were determined using eigenvalues >1.0, scree plot, pattern of loadings on each item (e.g. low loading), as well as the interpretability of factors. Oblique promax rotation was used to allow correlation between factors and factor loading cut-off was set as >0.3, which was similar to the factor loading cut-off used by the original developer of the scale (Annear et al., 2015; Tabachnick & Fidell, 2013). Loadings greater than 0.3 was used as the cut-off to enable the maximum number of items to be retained – since the DKAS was developed to assess knowledge, it would be optimal to retain as many items as possible – and also because this allowed a much better interpretation of solutions as compared to using an arbitrary 0.4 cut-off. Since we tried to retain the maximum number of items of the scale in the analysis, items with double-loading were also kept in the final scale. Finally, factor structure was formulated by retaining only items with largest loading score in each respective factor, taking into account the interpretability of solutions. The items with double loading were retained in the factors where they had higher loading on, as it helped to improve the interpretability of solutions without causing the factor to deviate from its supposed construct.

To understand the differences between caregiver’s sociodemographic characteristics and knowledge and understanding towards dementia, multiple linear regression was carried out to explore the correlates between sociodemographic characteristics and factor scores. Factor scores were tabulated by summing up the total scores of all the items in the respective factors, with higher scores representing better understanding towards dementia.

CFA was conducted using the ‘lavaan’ package under R program and EFA was conducted using the statistical software Mplus version 7.0. Both CFA and EFA had adjusted for categorical variables with the estimator of ‘Weighted Least Square Means and Variance Adjusted (WLSMV)’ (Beauducel & Herzberg, 2006). Descriptive analysis and multivariate linear regression were performed using IBM SPSS version 23 (IBM Corp. Armonk, New York. United States), with two-sided p value below .05 being considered as statistically significant.

Results

Descriptive analysis

The demographic characteristics of our sample are shown in Table 2. Participants had an average age of 55.6 with standard deviation (SD) of 11.8 years with majority being female caregivers (75.2%). Chinese accounted for more than three quarter of the participants (83%), followed by Malays (10.3%), Indian and others (6.7%). In terms of highest education level attained, 31.6% of participants had completed at least undergraduate education, 25.9% had an Institute of Technical Education (ITE), Polytechnic or Junior college (A level) education and 42.6% were secondary school’s General Certificate of Education (N/O level) graduates or below. Almost half of the participants were unemployed at time of assessment (42.9%), and majority (85.1%) had a salary of lower than SGD 6000 (the median monthly income in Singapore is SGD 4563 (Ministry of Manpower, 2020, January 30), approximately 3263 USD). Seventy-two percent (72%) of participants were ever married. In terms of caregiver’s relationship to care recipient, 15.2% were spouse, 17.0% of caregivers were son, 55.3% were daughter, while 12.4% were neither a spouse nor a child to the care-recipient (for instance, parent, sibling, niece, nephew, friend or in-law). Lastly, the average duration of caregiving history was 48.2 (SD = 50.3) months.

Table 2. Sociodemographic characteristics of participants (n = 282).

		n	%
Gender	Male	70	24.8
	Female	212	75.2
Ethnicity	Indian and others	19	6.7
	Malay	29	10.3
	Chinese	234	83.0
Education	N/O Level and below	120	42.6
	A level, ITE, poly	73	25.9
	University or above	89	31.6
Employment	Unemployed	121	42.9
	Employed	161	57.1
Marital status	Never married	79	28.0
	Ever married	203	72.0
Personal monthly income	Less than 2k SGD	56	19.9
	Between 2k and 6k SGD	78	27.7
	6k SGD or above	42	14.9
	NA	106	37.6
Relationship to patient	Others	35	12.4
	Son	48	17.0
	Daughter	156	55.3
	Spouse	43	15.2
		Mean	SD
Age		55.6	11.8
Duration of caregiving (months)		48.2	50.3

Building of DKAS 3-factor model

All 27 items of DKAS were ranked as ordinal data (2, 1 and 0). After conducting an EFA, the eigenvalues and scree plot on all 27 items of the DKAS suggested that 3-, 4- or 5-factor models were all potential solutions, and factor loadings in all three models were explored. Removal of items was based on the following priority: (1) item consistently showed lowest loading across all models; (2) item consistently cross-loaded across all models and (3) lowest loading. Eventually, a 3-factor model containing 23 items with an acceptable model fit indices was identified; χ²(df) = 282.22(187), CFI = 0.93, TLI = 0.91, RMSEA = 0.043. Removed items are documented in S1 Appendix.

The first factor consisting of 11 items was labelled as ‘misconceptions about dementia’ (MD), while the second factor comprising eight items described ‘caregiving considerations towards dementia’ (CD) and the last factor ‘dementia symptoms’ (DS) contained four items which are statements pertaining to dementia symptoms. Please refer to Table 3 for items loading in each factor. Composite reliability for the three factors was 0.79, 0.70 and 0.80, respectively. Higher summed score on a factor indicates a greater understanding towards that particular aspect of dementia knowledge. The summed scores for ‘MD’ was 10.35 (SD = 4.51, range 0–20, highest possible score = 22), 7.78 (SD = 3.47, range 0–16, highest possible score = 16) for ‘caregiving considerations towards dementia’ and 5.83 (SD = 2.29, range 0–8, highest possible score = 8) for ‘dementia symptoms’. The mean total score for all three factors combined was 24.07 (SD = 7.84, range 3–44, highest possible score = 46). The mean score for each factor (summed mean factor scores/number of items in factor) are 0.941 for ‘MD’, 0.984 for ‘CD’ and 1.459 for ‘DS’.

Table 3. Factor loading from EFA analysis on DKAS.

Item description	Factor loading
	f1	f2	f3
Factor 1 – Misconceptions about dementia
DKAS1 – Dementia is a normal part of the ageing process	0.355	–	–
DKAS 3 – People can recover from the most common forms of dementia	0.850	−0.371	–
DKAS 4 – Dementia DOES NOT result from physical changes in the brain	0.620	–	–
DKAS10 – Maintaining a healthy lifestyle DOES NOT reduce the risk of developing the most common forms of dementia	0.367	–	–
DKAS13 – Early diagnosis of dementia DOES NOT generally improve QOL for people experiencing the condition	0.392	0.327	–
DKAS14 – The sudden onset of cognitive problems is characteristic of common forms of dementia	0.343	–	–
DKAS15 – It is impossible to communicate with a person who has advanced dementia	0.412	–	–
DKAS17 – It is important to correct a person with dementia when they are confused	0.472	–	–
DKAS20 – People with dementia are unlikely to experience depression	0.657	–	–
DKAS21 – Medications are the most effective way of treating the behavioural symptoms of dementia	0.420	–	–
DKAS22 – People experiencing dementia DO NOT generally have problems making decisions	0.602	–	–
Factor 2 – Caregiving considerations towards dementia
DKAS5 – Most forms of dementia reduce the length of a person’s life	–	0.702	–
DKAS6 – Planning for end of life care is generally NOT NECESSARY following a diagnosis of dementia	0.325	0.352	–
DKAS8 – Most forms of dementia DO NOT generally shorten a person’s life	–	0.766	–
DKAS11 – Symptoms of depression can be mistaken for symptoms of dementia	–	0.351	–
DKAS12 – Exercise is generally beneficial for people experiencing dementia	–	0.414	–
DKAS18 – People experiencing advanced dementia often communicate through body language	–	0.457	0.327
DKAS19 – Uncharacteristic behaviours in a person experiencing dementia are generally a response to unmet needs	–	0.314	–
DKAS27 – Daily care for a person with advanced dementia is effective when it focuses on providing comfort	–	0.435	0.343
Factor 3 – Dementia symptoms
DKAS23 – Movement is generally affected in the later stages of dementia	–	–	0.679
DKAS24 – People with advanced dementia may have difficulty speaking	–	–	0.853
DKAS25 – People experiencing dementia often have difficulty learning new skills	–	–	0.676
DKAS26 – Difficulty eating and drinking generally occurs in the later stages of dementia	–	–	0.623

Multiple linear regression

Results from the multiple linear regressions showed that gender and education were both significant predictors of scores for MD and CD (refer to Table 4). Both factors showed similar findings in gender, in that being a male was associated with scoring lower as compared to being a female. However, with regards to education, those who had completed university education or more tended to score higher than their counterparts (N/O level and below, and ITE to Polytechnic) for MD. In terms of CD scores with regards to education, the only significant differences were between those who had completed N/O level and those who had minimally attained a degree. There were no significant association found between the scores of DS and caregiver’s sociodemographic characteristics.

Table 4. Sociodemographic correlates of three DKAS factors.

	Factor 1				Factor 2				Factor 3
Sociodemographic characteristics	β	95% CI		p Value	β	95% CI		p Value	β	95% CI		p Value
Gender
Male	−2.492	−4.625	−0.359	.022*	−1.941	−3.727	−0.154	.033*	−1.062	−2.253	0.129	.080
Female (Ref)	–	–		–	–	–		–	–	–		–
Ethnicity
Indian and others	−0.686	−2.688	1.316	.500	0.827	−0.850	2.504	.332	−0.173	−1.290	0.945	.761
Malay	−1.442	−3.116	0.232	.091	0.448	−0.954	1.851	.530	−0.790	−1.724	0.145	.098
Chinese (Ref)	–	–		–	–	–		–	–	–		–
Marital status
Never married	0.591	−0.537	1.719	.303	0.717	−0.228	1.662	−.137	0.289	−0.341	0.919	.368
Ever married (Ref)	–	–	–	–	–	–		–	–	–		–
Employment status
Unemployed	0.376	−1.445	2.198	.685	−0.273	−1.799	1.253	.725	0.272	−0.745	1.289	.599
Employed (Ref)	–	–	–	–	–	–		–	–	–		–
Personal income
<SGD 2000	0.169	−1.834	2.172	.868	−0.722	−2.400	0.956	.398	0.105	−1.223	1.013	.853
SGD 2000–SGD 5999	0.063	−1.959	2.086	.951	−0.850	−2.545	0.844	.324	−0.118	−1.247	1.011	.837
SGD 6000 or above	0.101	−2.313	2.516	.934	−0.893	−2.916	1.130	.386	0.058	−1.290	1.406	.932
Not applicable (Ref)	–	–	–	–	–	–		–	–	–		–
Relationship to care recipient
Others	1.851	−0.383	4.086	.104	0.674	−1.198	2.547	.479	−0.217	−1.465	1.031	.732
Son	2.071	−0.362	4.505	.095	0.725	−1.314	2.796	.484	0.277	−1.082	1.636	.689
Daughter	0.492	−1.516	2.499	.630	−0.295	−1.977	1.387	.730	−0.063	−1.185	1.058	.911
Spouse (Ref)	–	–	–	–	–	–		–	–	–		–
Highest education attained
N/O level and below	−3.601	−4.904	−2.297	.000**	−1.271	−2.363	−0.179	.023*	−0.432	−1.160	0.296	.243
ITE, A level or poly	−1.966	−3.303	−0.629	.004**	−0.341	−1.462	0.779	.549	−0.476	−1.222	0.271	.211
University or above (Ref)	–	–	–	–	–	–		–	–	–		–
Age	−0.023	−0.080	0.034	.429	0.008	−0.039	0.056	.730	0.017	−0.015	0.049	.296
Duration of caregiving (months)	0.004	−0.006	0.014	.424	0.005	−0.003	0.014	.196	0.000	−0.005	0.006	.873

* p < .05.

** p < .01.

Discussions

Contrary to both the factor structures proposed by earlier studies (Annear, Otani, et al., 2017; Annear, Toye, et al., 2017), our findings indicated a 3-factor structure, retaining 23 out of the original 27 items from the scale. Given that studies have found levels of dementia knowledge to vary due to sociodemographic differences (Ayalon, 2013; Gray et al., 2009), the dissimilarity between our factor structures and previous studies could be attributed to differences in the sample. In the earlier DKAS CFA study, the sample comprised an international cohort of participants who had completed an online dementia MOOC. Furthermore, majority of the participants recruited in aforementioned study were workers in healthcare settings (i.e. nurses and professional care worker) while only a small number were family caregivers (Annear et al., 2015). In contrast, our study specifically recruited only informal caregivers of dementia patients in Singapore. Hence, it is likely that participants in the earlier study were relatively more well-informed about dementia at the time of assessment as compared to our sample. Similarly, the participants in the Japanese DKAS study consisted of a cohort of health students, academics or professionals, which is unlike our sample that comprised of only informal caregivers.

Our proposed first factor is coined as ‘misconceptions about dementia’ because the items comprising this factor such as ‘Dementia is a normal part of the ageing process’, ‘It is important to correct a person with dementia when they are confused’ and ‘Medications are the most effective way of treating the behavioural symptoms of dementia’, represent some of the misperceptions that people may have towards dementia – if they do not possess an in-depth enough understanding towards the disease – possibly resulting in them imposing higher expectations on a dementia patient. For instance, misunderstanding that medications are the most effective way of treating dementia might mislead a caregiver to overestimate the efficacy of medication treatment. Thus, if patient’s condition does not improve upon medication treatment subsequently, this might lead to the caregiver experiencing greater subjective distress (Graham et al., 1997; Sörensen & Conwell, 2011) eventually. The two items with double loading that were retained in this factor, namely DKAS 3 and 13 also relate to misconceptions about dementia. For example, a caregiver with the misconception that people can recover from the most common forms of dementia (DKAS 3) may hold mismatched expectations towards the care recipient, thereby possibly leading to greater subjective distress. Likewise, if a caregiver or PWD has the misconception that early diagnosis of dementia does not generally improve QOL (DKAS 13), it may engender procrastination in seeking treatment due to the belief that there’s limited benefit in seeking medical advice given the deteriorative nature of dementia (Vernooij-Dassen et al., 2005). Hence, higher score on this factor signifies lower misperception towards dementia, and by extension, lower likelihood of over-expectations towards a dementia patient’s prognosis.

The second factor was labelled as ‘caregiving considerations towards dementia’ because three of the items (items 5, 6 and 8) loaded here relate to life expectancies, which are important for family caregivers for proceeding with medical decisions and re-evaluating caregiving goals. The other five items that loaded on this factor describe important information that would facilitate quality of care towards the dementia patient. For instance, item 11 helps to assess whether one is aware that ‘symptoms of depression can be mistaken for symptoms of dementia’, which is important because depression is a relatively common occurrence amongst dementia patients (30%) (Enache, Winblad, & Aarsland, 2011). Unfortunately, symptoms of depression can often be mistaken as symptoms of dementia (SCIE, 2015), which results in this comorbid condition being undiagnosed and untreated amongst dementia patient, subsequently impairing both the patient’s and caregiver’s quality of life (Enache et al., 2011). The three items with double-loading that were retained in this factor are DKAS 6, 18 and 27. Like DKAS 6, DKAS 18 and 27 also relate to caregiving considerations. Arguably, if a caregiver understands that their care recipient tends to communicate through body language more (DKAS 18), they may be able to better infer the needs of the care recipient. Similarly, the quality of care given to the care recipient might be more effective if the caregiver understands that emphasizing on comfort is important for advanced stage dementia (DKAS 27).

The third factor was termed as ‘dementia symptoms’, as all the items that loaded on to this factor such as ‘people experiencing dementia often have difficulty learning new skills’ relates to common observable behavioural symptoms that dementia patients typically exhibit.

Our proposed 23-item 3-factor model contains domains that are dissimilar from the 25-item 4-factor model proposed by the scale developer (Annear, Toye, et al., 2017). Although one of our domains (caregiving considerations towards dementia) has a similar name to one of the domains (care considerations) in the 4-factor model, the items that load into the two respective domains are quite different. Nonetheless, despite the disparity in factor structure, the constituent items from the three domains in our factor structure still cover extensive information pertaining to dementia such as pathology, symptomatology, risks and health promoting factors, communications and care considerations. In comparison to the 18-item unidimensional Japanese DKAS model (Annear, Otani, et al., 2017), our 23-item 3-factor model arguably assesses the dementia knowledge of informal caregivers more extensively, with the added advantage of allowing subscale evaluation or comparison.

As mentioned in the ‘Results’ section, the maximum summed factor scores attainable from the new factor model consisting of 23-item is 46, and mean score of the participants was 24.07. Based on this finding, it is recommended that knowledge of dementia should be improved in informal caregivers in Singapore, given the benefits entailed. This signals the need for interventions targeted at dementia caregivers in Singapore to also impart knowledge on dementia, besides teaching coping techniques and offering psychological comfort. Alternatively, given how immensely taxing and time-consuming caregiving can be, another suitable way for caregivers to improve their knowledge is to undertake online dementia course that have been documented to be effective at elevating understanding towards dementia (Annear et al., 2015; Eccleston et al., 2019; King et al., 2014).

Comparison of mean score between factors shows that participants scored better on DS (1.459) than MD (0.941) and CD (0.984) which have somewhat similar scores. This suggests that being informed about symptoms of dementia does not necessitate better understanding of the needs of dementia patients nor alleviates the misconceptions that one may have towards dementia. Furthermore, it implies that interventions targeted at dementia caregivers should place more emphasis on the caregiving considerations and needs of dementia patients, in order to improve quality of care provided by caregivers (Chodosh et al., 2007; Gibbons et al., 2005; Matsuda et al., 2018). These interventions should also seek to inform caregivers on the non-symptomatic traits of dementia so as to alleviate caregivers subjective burden which may arise from mismatched expectations (Chiu et al., 2013; Sörensen & Conwell, 2011).

The analysis of sociodemographic correlates with our three factors suggests that a being a male caregiver is significantly associated with lower scores on factors MD and CD, and this is concordant with previous studies that examined correlates of dementia knowledge (Arai, Arai, & Zarit, 2008; Sun, Gao, Shen, & Burnette, 2014; Werner, Goldberg, Mandel, & Korczyn, 2013). Low and Anstey (2009) found that females in general were more worried than males about getting dementia (Low & Anstey, 2009). As having a close relative with dementia means increased risk of oneself developing dementia and women have a higher incidence rate of dementia than man, female caregivers may therefore have increased awareness of the risk of contracting dementia themselves (Alzheimer’s Society, 2016). Thus, this may motivate female caregivers to learn more about the condition, possibly explaining the gender difference with regards to knowledge. Alternatively, this may be linked to how male caregivers have weaker support networks and are also less likely to seek out programs that can bolster their caregiving capabilities (Lopez-Anuarbe & Kohli, 2019). In which case, their knowledge may be implicated insomuch that even if they encounter any issues with caregiving or experience any knowledge gaps in caregiving, they have less revenues to seek help, ask questions or attain appropriate advice from. Another postulation is that because of traditional masculinity, even if male caregivers had encountered difficulties in caregiving, they might be reluctant to disclose and seek guidance for it, as doing so would imply weakness or not being ‘man enough’ to deal with the job (Baker, Robertson, & Connelly, 2010). This thus hinders their capability to gain more knowledge about dementia and its caregiving.

An interesting finding from our study is the absence of sociodemographic predictors for our third factor ‘DS’, which might be attributable to the fact that items pertaining to symptoms of dementia might be more obvious than items loaded in the other two factors. A systematic review on public’s knowledge towards dementia found that across 17 studies reviewed, despite participant’s overall knowledge of dementia being poor, their awareness of dementia symptoms was good (Cahill, Pierce, Werner, Darley, & Bobersky, 2015). Hence, it can be extrapolated that generally speaking, people tend to be relatively knowledgeable towards symptoms of dementia, even though they might not be as well-informed towards other facets of dementia. Since our participants consisted only of primary caregivers of dementia patients, they are even more likely than the general public to be aware of dementia symptoms, having witnessed the symptoms first-hand during their involvement with their care recipient. This might account for a better understanding of dementia’s symptoms, resulting in a lack of variation in DS scores, which attributed to the insignificance in findings.

Being of lower educational levels was associated with poorer scores on factor 1 and 2, which is congruent with findings from previous studies that evinces the positive association between levels of education and dementia knowledge (Ayalon, 2013; Edwards, Cherry, & Peterson, 2000; Werner, 2001). While there were significant differences in MD scores between participants with lower education levels in comparison to university or above participants in CD scores, there were only significant differences between those with N/O level and below versus university and higher education participants. A plausible explanation for this outcome is that many items in MD are comparatively more related to pre-dementia (risk prevention, onset and causes of dementia), while all the items in CD describes statements strictly pertaining to dementia, which are more related to care needs of dementia patient. Thus, even though education levels may influence dementia knowledge, it may not have as big an impact on the scores of CD insomuch that participants’ personal experience in taking care of a person with dementia might have positively mediated their knowledge on CD. In contrast, participants’ personal caregiving experience is unlikely to improve their knowledge on MD, unless they specifically make an effort to acquire more knowledge about dementia. Existing literature on the relationship between health literacy and educational level generally evince a positive association between the two variable (Paasche-Orlow, Parker, Gazmararian, Nielsen-Bohlman, & Rudd, 2005), possibly because more highly educated people have greater accessibility to health information.

Lastly, one notable finding is that duration of caregiving (months) did not significantly affect a caregiver’s knowledge towards dementia across all three factors, insinuating that experience in caregiving does not necessarily entail better understanding of dementia. A plausibility is that perhaps the knowledge gained from caring for a person with dementia is very minimal, or that the gain in knowledge acquired from caregiving plateaus after a certain period. On this note, it is worth mentioning that some studies have elucidated that longer duration of caregiving is associated with greater caregivers’ psychological distress (D’Onofrio et al., 2015; Heok & Li, 1997; Torti, Gwyther, Reed, Friedman, & Schulman, 2004; Tremont, 2011). Since knowledge of dementia is evinced to be a protective factor against caregiver’s subjective distress (Graham et al., 1997; Sörensen & Conwell, 2011; Werner, 2001), this finding reinforces the need for caregivers to improve their understanding towards dementia even if they have been a caregiver for a prolonged period of time.

Limitations

Firstly, our study, being cross-sectional, precludes us from drawing any causal conclusions from our findings. Secondly, given the convenience sampling strategy adopted for the recruiting of participants in this study, the factor structure obtained might not be generalizable beyond informal caregivers. Lastly, given the comparatively smaller sample size for our EFA, it calls into question the stability of our solutions. However, some scholars have endorsed that a minimal of 200 sample suffices for EFA (Guadagnoli & Velicer, 1988; Matsunaga, 2010) and our sample size was indeed higher than 200. In addition, our factor analysis was run on a relatively high subject to item ratio (approximately 10:1), generated a solution with a relatively small number of factors (three factors), with each of the factors containing at least more than three items, all of which lends credence to the stability of this factor structure (Costello & Osborne, 2005). Regardless, given the comparatively smaller sample size, there is a need to examine the validity of the reported DKAS structure model by CFA in future studies with larger samples.

Notwithstanding these limitations, this is the first study that has examined the factor structure of the DKAS amongst a sample of informal caregivers in Southeast Asia. A different factor structure was identified in our Singapore sample, indicating that cultural differences and caregiver’s relationship to dementia patient (caregiver vs. non-caregivers) might influence one’s knowledge towards dementia. Ultimately, more studies need to be carried out to verify the validity of our 3-factor model, as well as the generalizability of the model for assessing other groups of individuals.

Conclusion

To our knowledge, this is the first study that examined the appropriateness of three different factor models of the DKAS amongst a sample of informal caregivers in Singapore. Both the 4-factor and unidimensional models (Annear, Otani, et al., 2017; Annear, Toye, et al., 2017) demonstrated poor fit, and a new 23-item 3-factor model was proposed after performing EFA. Results from this study suggest a poor overall understanding towards dementia amongst the participants. Sociodemographic correlates with the new factor structure were also investigated. Being a male and having lower education levels were associated with poorer scores on MD and CD domains. In addition, duration of caregiving had no significant effect on dementia knowledge. Since knowledge facilitates the care management of dementia patients, our findings highlight the need for more dementia educational intervention tailored for caregivers, and more specifically male caregivers and those of lower educational level. Future studies could explore the validity of our contemporary factor model, and whether it could be generalized for use beyond caregivers.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data sharing statement

Readers who wish to gain access to the data can write to the senior author MS @ [email protected] to request access.

Additional information

Funding

This project was supported by funding from the National Medical Research Council in Singapore. The study was funded by Singapore Ministry of Health’s National Medical Research Council under the Centre Grant Programme (Grant No.: NMRC/CG/004/2013) and the Institute of Mental Health Bridging Fund (CRC Ref: 545-2016). The funding source had no role in the study design and in collection, analyses and interpretation of data and in writing this report.

Notes on contributors

Gregory Tee Hng Tan

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Qi Yuan

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Fiona Devi

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Peizhi Wang

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Li Ling Ng

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Richard Goveas

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Siow Ann Chong

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.

Mythily Subramaniam

Tan: recruitment, data collection, data cleaning and analysis, preparation of manuscript. Yuan: design and development of study, data cleaning and analysis, critical feedback on draft manuscript. Devi & Wang: recruitment, data collection, data cleaning, and feedback on study design and draft manuscript. Ng & Goveas: referral of participants and feedback on draft manuscript. Chong & Subramaniam: review and feedback on study design, critical feedback on draft manuscript.