Sample

The study design was a cross sectional survey. A total of 152 PWD living at home and their 128 informal carers were recruited from health and social services, and voluntary organisations in North East London (65 PWD and 54 carers), Cambridgeshire (81 PWD and 68 carers) and Liverpool (6 PWD and 6 carers). Inclusion criteria for PWD were: being aged 60 or over, having a diagnosis of dementia according to DSM-IV and living at home (not in institutions). A person was considered an informal carer (spouse, relative, friend or neighbour) when he/she was knowledgeable about the person with dementia and spent a minimum of 4 hours a week in direct contact (face to face) with them. Only one person was considered the nominated carer. A sample, similar to those obtained from previous community studies of dementia in terms of living situation, was recruited: 65% of participants living with another person and 35% living alone [9, 10].

At each centre, the manager or appropriate member of staff was requested to make a first approach either with the PWD's carers or the PWD themselves (depending on dementia severity and/or living situation) to give them the Information Sheet and to discuss if they were willing to be approached by the researcher regarding this study. Participants who had no objection were contacted by the researcher by phone and were given more information about the study as required. In addition, details of people who had attended the centre (PWD's name, carer's name, address, phone, etc.) were provided and a letter was sent to the carer and/or patient including information sheets about the study. One week after, the researcher contacted them by phone, answered any questions they might have, explained the study and looked for their willingness to participate. If the potential participant agreed to be involved, either the researcher or an interviewer arranged a day to carry out the interview at their homes.

Procedure

The study was carried out in accordance with the latest version of the Declaration of Helsinki. Ethical approval was granted by East London & the City HA Research Ethics Committee 3. All participants gave informed consent. Once the interviewer was at the PWD's or carer's home, they answered any further queries and sought informed consent as follows: written consent by signing a Consent Form was required from people with dementia and their caregivers. In order to achieve this, the interviewer approached potential subjects to explain the study and to inform them about their right to withdraw at anytime. Some PWD (11, 7.3%) were unable to provide written consent. When this occurred, the interviewer sought their assent (verbal consent). During this process, the interviewer made sure that he/she had taken as much time and care in explaining the information about this research as simply as possible. The interviewer avoided using long sentences and attempted to reduce any distractions. To find out if the participants have understood the information given, the interviewer observed their ability to ask any relevant questions and also requested the participant to repeat back the information and how it would relate to them. In addition, the interviewer clarified any doubts about the study and reiterated their right to withdraw at anytime.

Interviews were carried out by an experienced clinical psychologist and old age psychiatrists who were trained to undertake the interview. All the interviews with PWD were carried out at their homes. Interviews with carers were undertaken either at the PWD's home, their own home or at a health centre (e.g. memory clinic, day hospital). Whenever possible, the PWD and the carer were interviewed separately. However, some carers wanted to be present during the PWD's interview. In this situation, the interviewer emphasized the fact that he/she was interested in knowing both parties' views about the PWD's needs and quality of life; and, as they were unique human beings, both opinions do not necessarily have to be coincident and both were considered valid. Interviews were terminated immediately in presence of any sign of discomfort. In addition, if the conditions at home were not appropriate to undertake the interview with the carer, the interviewer arranged a next meeting in other location.

Using standardised instruments, PWD were interviewed about their needs, cognitive status and quality of life. This interview took an average of 30 minutes depending on dementia severity. Carers were interviewed about the PWD's sociodemographic details, needs, behavioural and psychological symptoms, functional status, quality of life, social networks, and services received. In addition, carers were assessed about their own sociodemographic details, depression, anxiety and burden. This interview took about two hours and a half.

Measures

Data Analysis

Statistical analyses were undertaken using the SPSS 15.0 software package and AMOS 7.0 [31]. The significance level used was p < 0.05. Since needs (met and unmet) were not distributed normally, non-parametric tests were performed. For a better understanding of the results, when comparing groups, means instead of ranks are shown.

In order to identify predictors of unmet need, a stepwise multiple linear regression was performed. In addition, in order to test the theoretical model a path analysis was carried out using AMOS 7.0. For this analysis, only the main variables of this study were considered: living situation (alone vs. with others), behavioural and psychological symptoms (NPI score), services received, unmet needs, social networks, quality of life (rated by patients and carers) and carers' mental health (depression, anxiety and burden). Variables skewed ≥1 ('Unmet needs' and 'NPI Score') were transformed [32]. Maximum likelihood method was used to estimate the model. Chi-square statistic for the model was reported. A non significant χ² value indicates that the model does not occur by chance. The Normed Fit Index (NFI), and Comparative Fit Index (CFI) were reported. Values ≥ 0.90 indicate good fitting of the model [33]. Finally, the Root Mean Square Error of Approximation (RMSEA) which is sensitive to the number of estimated parameters in the model and penalizes the lack of parsimony was reported. A RMSEA value less than 0.05 with a narrow confidence interval (CI) denotes adequate parsimony [34].

Participants

Needs

Factors associated with unmet needs

Predictors of unmet needs

Relationship between unmet needs, social networks and quality of life

In the first model, three regression analyses were performed; one in which social network was entered as the independent variable (IV) and quality of life rated by carers as the DV, another in which social network was treated as the IV and unmet needs (the presumed mediator) was treated as DV, and a last one in which social network and unmet needs were entered as IV with quality of life rated by carers as the DV. The amount of mediation (indirect effect) was calculated by subtracting the regression coefficient (social network to quality of life rated by carers) in the third regression (with unmet needs controlled) from the regression coefficient (social network to quality of life rated by carers, again) in the first regression (with unmet needs not controlled). The reduction in the regression coefficient from the first regression to the third regression when unmet needs was controlled was 0.537 - 0.200 = 0.337 which suggests a partial mediating effect. The Sobel test indicated that the direct effect of social network on quality of life rated by carers was significantly reduced when unmet needs was added to the equation Z = 2.11, p < 0.05. The same procedure was followed to determine if unmet needs acted as mediator between social networks and self-rated quality of life. The reduction in the regression coefficient from the first regression to the third regression when unmet needs was controlled was 0.965 - 0.767 = 0.198 which suggests a partial mediating effect. The Sobel test indicated that unmet needs does not significantly mediate the relationship between social network and self-rated quality of life (Z = 1.45, p = 0.14).

Testing the theoretical model using path analysis

The overall goodness of fit, for the hypothesized model, given by chi-square was χ² = 43.1, df = 17, p < 0.001. NFI was 0.67 and CFI was 0.73 indicating that the hypothesized model did not fit the data. After this result, an exploratory approach was adopted in order to construct a model that better reflects the data. The original model was modified following this process: Step 1: All the paths that had significant correlations were added to the hypothesized model. Regression paths were added from behavioural and psychological symptoms to unmet needs, from behavioural and psychological symptoms to carer anxiety (which was the only carer mental health factor significantly associated with unmet needs), from carer anxiety to unmet needs, from carer anxiety to quality of life rated by carers, from carer age (which in bivariate analyses was found associated with unmet needs) to unmet needs, from self-rated quality of life to services received, and the last one from living with others to social network. Two covariance paths (drawn from bivariate analyses) were also added, one between carer age and living with others and the second one between carer age and behavioural and psychological symptoms. The fit of the model after Step 1 was χ² = 9.28, df = 17, p = 0.66; NFI = 0.86; CFI = 1.0; RMSEA = 0.0, indicating that this model still did not fit the data. In order to obtain a model which fits the data, the next steps (2 and 3) consisted of removing non significant paths one by one, starting with the least significant. Step 2: The path from carer age to unmet needs was removed (χ² = 10.24, df = 18, p = 0.74; NFI = 0.90; CFI = 1.0; RMSEA = 0.0). Step 3: The path from unmet needs to quality of life rated by carers was removed. After removing this last path, the rise in χ² statistic for the model was more than the critical ratio (showing that the resulting model is significantly less good at fitting the data). The fit of the new model improved considerably (χ² = 11.79, df = 19, p = 0.89; NFI = 0.92; CFI = 1.0; RMSEA = 0.0), however, in order to obtain a more parsimonious model, the accepted criterion of removing one by one the paths which were not significant --starting with the least significant-- until the point when the rise in χ² statistic for the model was more than the critical ratio (showing that the resulting model is significantly less good at fitting the data) was used. Figure 1 shows the model that best fit the data (n = 152). The general goodness of fit was χ² = 13.4, df = 22, p = 0.92 indicating a good fit. NFI was 0.92 and CFI was 1.0 showing that the model fit was very good. Finally, the RMSEA index was 0.00, 90% CI (0.00-0.02) indicating also a parsimonious model. Having more behavioural and psychological symptoms (as measured by NPI), living alone and having a low-community involvement network were significant predictors of unmet needs. Higher levels of anxiety in carers were predicted by having more behavioural and psychological symptoms in PWD. A higher carer-rated quality of life was predicted by a higher self-rated quality of life and fewer behavioural and psychological symptoms. In addition, a higher self-rated quality of life was predicted by fewer unmet needs; and a higher self-rated quality of life was associated with a low use of services. Finally, behavioural and psychological symptoms, living situation (alone vs. with others), and social networks, all had indirect effects on self-rated quality of life through unmet needs, and also on carer-rated quality of life through self-rated quality of life.

Factors associated with unmet needs

The mean number of unmet needs of PWD in this research was higher than those assessed in sheltered housing [8] but it was lower than the unmet needs found for PWD in residential care [1]. This is congruent with the fact that whereas PWD living in sheltered housing are less dependent and might have better access to support, PWD in residential care had a higher prevalence of depression, anxiety, physical dependence, and behavioural problems [1]. Despite this, the most common unmet needs were the same as in residential care suggesting that PWD may have similar unmet needs in the community or in institutional care. The finding that PWD who had a low-community involvement network type had higher unmet needs than those who had a high-community involvement suggests that it is important to maintain community involvement for PWD for as long as possible. In the UK, the resources available for PWD and their carers have increased over the last years, although this is not necessarily synonymous with better coordination and appropriateness in the delivery of care [36]. Furthermore, for PWD who do not like to attend any kind of group, there are befriending services and outreach workers which provide a good opportunity to socialise, however the current availability of these in the UK is limited. The contact of PWD with the world outside the home (even if the person is visited at home) could be beneficial in several ways: increasing their involvement with the community, diminishing their psychological and social unmet needs and ultimately improving their quality of life.

The total number of services used by PWD was not associated with unmet needs. However, PWD who had attended a day hospital or a day centre over the last three months had fewer unmet needs than those who had not. Perhaps these services are more suitable for the needs of people living at home because they can offer support for several areas such as daytime activities, memory, and company. However, these services were not within the most used by PWD. There could be several reasons for this, such as the lack of need, lack of availability, lack of knowledge about their existence, lack of suitability for the carer, etc. In addition, PWD who had been visited by a social worker and those who had received home delivered meals had higher number of unmet needs. However, after controlling for living situation, this later association was explained by the fact that people who used home delivered meals were living alone [37].

Higher behavioural and psychological symptoms (NPI score) were associated with higher unmet needs. In line with Hancock's care home study [1], people with depression and dementia had significantly more unmet needs. Neil & Bowie [38] found that depression, agitation and anxiety were reported by family carers as the most distressing symptoms of dementia. Since the family is the main provider of support for the needs of PWD living at home and behavioural and psychological symptoms are highly correlated with carer stress and burden [39] it is not surprising that people having more behavioural and psychological symptoms were found to have significantly more unmet needs.

Consistently with previous studies, carers' ratings of quality of life were lower than those of PWD [3, 18]. In the present study PWD with higher self-rated quality of life had fewer unmet needs, but this association was not significant. In addition, higher carer defined quality of life among PWD was associated with fewer unmet needs. In residential care, Hoe et al. [3] found that higher self-rated quality of life correlated with fewer unmet needs rated by researchers. In residential care, researcher ratings of unmet needs may be more likely to mirror PWD's views about their own quality of life due to the homogeneous environment for participants. In contrast, PWD living at home will have a more varied physical and social environment (e.g, activities, relationships, support) and so their own views of quality of life may well differ from the researchers assessment of needs.

Having high behavioural and psychological problems, a low-community involvement network, being cared for by a younger carer, and being looked after by an anxious carer explained half of the variability in unmet needs. Besides carer's age, all of these factors are modifiable. Thus, changing these factors would provide a chance to diminish the number of unmet needs. Psychosocial interventions and/or pharmacological interventions could be useful in diminishing neuropsychiatric symptoms [40]. In addition, Cooper et al. [41] have pointed out that psychosocial interventions aiming to diminish dysfunctional and increase emotion-focused coping could reduce carer's anxiety. The not explained variance in unmet needs may be caused, among other factors, by individual differences. The CANE allows the care practitioner to identify these individual unmet needs making possible the referral of each client to suitable interventions.

The mediation analyses showed that unmet needs were partially mediating the relationship between social networks and quality of life (self-rated or rated by carers); however the mediation effect was significant only for carer-rated quality of life. The PANT measures social networks from a structural perspective but does not provide any information about perceived social support, this is, the level of satisfaction with the help provided by the members of their social network. Thus, it might be that when PWD assess their quality of life, their perception of the quality of the support received may be more important than the structural aspect of their social network which may in turn perhaps be considered more by the carers when assessing the person with dementia's quality of life.

Path analysis

The hypothesised model did not fit the data but an alternative model generated using the path analysis technique showed that people who had more developed social networks had significantly fewer unmet needs and, at the same time, those with fewer unmet needs had significantly higher self-rated quality of life. Thus, when rating their quality of life, PWD give more importance to the number of areas in which they are not receiving appropriate support (unmet needs) rather than to the structural aspects of their social network, even though these two variables are significantly associated with each other. Unmet needs may be reflecting the quality of care received which may have an impact on how PWD rate their own quality of life.

Improving the interventions and quality of care for dementia are two of the main issues highlighted by the National Dementia Strategy in the UK [42]. Therefore, the findings of this study have important implications for service providers. Behavioural and psychological symptoms, living situation, and social networks were associated with unmet needs and these are potentially modifiable. There are interventions available to reduce behavioural and psychological symptoms and to increase PWD's community involvement. Since changing living situation is difficult (it is unlikely that people living alone will shift to live with others, and that may not be necessarily beneficial), care providers should ensure that PWD living alone have the opportunity to receive any services that allow them to be in contact with the community, either by attending a psychosocial intervention (e.g. cognitive stimulation or reminiscence therapy) or social group; or by receiving support from a befriending visitor or outreach worker. Given their relevance to practice, it may be surprising that formalised assessment of social networks has not become part of case management. This could reflect reluctance on the part of practitioners to introduce standardised measures perhaps because it may be seen as increasing their workload and/or of little relevance to their everyday work.

Methodological Considerations

Since the severity of the cognitive impairment could make difficult to contact PWD directly, whenever possible, the carer was also approached (either by phone or letter). However, this strategy resulted into two problems (sources of recruitment bias). First, in some cases the carer refused to participate on behalf of the person with dementia and consequently the person with dementia did not have the opportunity to be seen and asked by the researcher about their willingness to take part in the study. Second (either in absence of a carer or when the carer could not be reached) some PWD living alone who were not able to follow the conversation by phone and consequently could not agree to be interviewed, had to be excluded.

Even though carers were present in less than 10% of the interviews, PWD in this situation may have felt that they could not express their views freely, and consequently their answers may have been affected by the presence of the carer.

Since the interviews were long, the carers sometimes got tired and were unable to continue answering questions. Consequently, the last instrument (NPI) was sometimes not administered in order to prevent fatigue. The average interview took between two and a half to three hours. Interviews were carried out at the participant's own home (not in a clinical setting) which probably was more comfortable for the participants. If researchers had used the time only to get specific and brief answers to the questions, the interview would probably have been shorter. However, doing this would have not contributed to establish a good rapport with the person with dementia and carers. When carers got tired, it was useful to have a break and continue the interview after that. However, because of limited financial resources, it was generally not possible to visit people twice. In future research the interviews could be carried out in two sessions. Regarding the instruments, some of them could be replaced or omitted, or short versions could be used. For example, since social networks and use of services was measured by the PANT and CSRI respectively, it would be feasible to use only the short version of the CANE. Also, the CSRI could be shortened including only those services that are commonly used by PWD living in the community; however this could lead to problems estimating the overall costs.

Limitations

In the regression analysis performed to determine which variables were the best predictors of unmet needs, "living situation" did not result in a significant predictor whereas in the path analysis the same variable came up as a significant predictor of unmet needs. This discrepancy may have been the result of the different sample size used for each analysis. While the regression was performed using only those participants who had complete data for all the predictors included in the model (n = 95), the path analysis was performed using maximum likelihood method, which dealt with missing data generating a model including the 152 participants. In addition, since a broad definition of carer was used, distinctions between those without a carer and those living with alone with an available carer were limited. Also, the regression model may have particularly excluded some people living alone who had no carer available. Therefore, the model generated using the path analysis should be considered as a better reflection of the relationship between living situation and unmet needs, indicating that living situation acts as a significant predictor of unmet needs. Higher behavioural and psychological symptoms (NPI score) were associated with a higher unmet needs but some of these symptoms assessed by the NPI are also included in the CANE (e.g. behaviour, psychological distress) which could have confounded these associations. However, the most frequent unmet needs of PWD (except for psychological distress) were found in areas not considered in the NPI such as, daytime activities, company, eyesight/hearing and accidental self-harm. The fact that only two people from minority ethnic groups was recruited, may be reflecting the difficulties these groups have in getting information about service availability and in accessing services in the areas where the recruitment was undertaken. Since this was a cross-sectional study, causality cannot be established from the results. Also, the final model generated from the path analysis was produced using an exploratory approach, so it has to be considered as a hypothetical model which needs to be tested in future research.