A longitudinal exploration of mental health resilience, cognitive impairment and loneliness

There is a growing interest in how people living with dementia may achieve good outcomes and be resilient despite their health challenges. Understanding what might be important for resilience in this population is largely untested theory.


| INTRODUCTION
In older age, awareness of changes in cognitive function which may presage the onset of dementia can be extremely stressful. Cognitive impairment is associated with a significantly higher risk of experiencing depression and anxiety, 1 and decreasing cognitive function is also a risk factor for self-reported loneliness. 2,3 Loneliness itself may lead to faster rates of cognitive decline 4 and dementia has been described as 'the hidden voice of loneliness'. 5 'Loneliness', defined as a negative emotional state arising from dissatisfaction with the quantity and quality of social resources, 3 is related to a wide range of health outcomes. 6 Contemporary policy now recognises loneliness as a major public health issue, and in 2018, the United Kingdom became the first country in Europe to appoint a Minister for Loneliness. The public health impacts of loneliness in the context of cognitive impairment are compounded by the predicted increase in the number of people with a dementia 7 and add to dementia's position as a global health challenge and an international public health priority. 8 Whilst cognitive impairment may correlate with poor outcomes, not all individuals will be affected in the same way. This paper focusses on those who do not appear to experience adverse consequences, despite lower cognitive function, and asks whether these individuals, who may be described as resilient, are also less likely to experience loneliness. There is a growing interest generally in people who appear resilient, who despite health challenges, do not experience adverse consequences.
How best to measure resilience is controversial. A number of resilience measurement scales exist; most measure factors that facilitate a resilient outcome, focussing mainly on psychological aspects. 9,10 Contemporary research recognises that an evaluation of resilience should take into account both the adversity and the outcome of interest, and that assets and resources both within the individual and within their social context are important for enabling a good outcome despite adversity. 11 This is often described as an ecological model of resilience. 12 For example, Joling et al. 13 examined resilience in dementia caregivers, with resilience operationalised as low reported levels of psychological distress despite facing substantial care demands (e.g., caring for a relative with more severe dementia, or self-care limitations). Using data from four different studies, the proportion of people who could be defined as resilient was ascertained and the determinants of resilience were explored. The study demonstrated the utility of a measurement approach more closely tied to a conceptual understanding of resilience.
A review of longitudinal resilience studies notes most research using cohort data operationalises resilience as the absence of psychiatric distress (e.g., no depression; no anxiety) in the face of an adversity, 14 defined more precisely as 'mental health resilience' (MHR). 15 However, Cosco et al. 14 noted that measures of positive function (e.g., well-being) had not been used, yet well-being may not be adequately reflected simply by the absence of psychiatric symptoms. A wide range of adverse circumstances are described by Cosco et al. 14 (e.g., bereavement), but no studies in their review examined the experience of cognitive impairment. In response, this paper builds on the approach taken by Joling et al. 13 and the recommendations of Cosco et al. 14 in order to provide a rigorous evaluation of resilience in the context of cognitive impairment and possible protection against loneliness.
A holistic assessment of resilience requires the identification of environmental and individual aspects commonly referred to as 'protective factors' or the 'resilience reserve'. Recently, there has been a shift towards thinking about how people with dementia might achieve positive outcomes and resilience. 16 However, there is little published research on resilience and dementia from this perspective.
Other factors important for resilience can be hypothesised from the broader literature. There is good evidence for some of the important and potentially modifiable lifestyle risk factors such as social engagement, physical activity, diet and alcohol consumption 17 that may impact on cognitive decline, Alzheimer's disease and other dementias. These factors are also recommended for good mental health. 18 Subjective memory complaints, which may reflect awareness of cognitive difficulties, have been associated with anxiety and depression for people living with cognitive impairment. 19 Exploring the role of these factors may offer some preliminary indication of relevant assets and resources.
The present study accordingly aims to contribute new insights into resilience and cognitive impairment, building on developments in resilience measurement and a recognition of the potential for positive responses to cognitive impairment and dementia. Reflecting suggestions regarding the operationalisation of resilience 11,14 we explore, for the first time in people with cognitive impairment, resilience conceptualised as the absence of psychological distress (no depression or anxiety), together with the presence of well-being (defined as MHR). The objectives are as follows: � To identify participants with good mental health over time, despite cognitive impairment (MHR).
� To explore some of the factors that may enable MHR.
� To examine whether MHR is a determinant of loneliness, as a key public health outcome, over time. They were followed up 2 years later (N = 2236). The response rate in wave 1 was 44%, and in wave 2, 70%.

| Data source and participants
Ethical approval was granted by the appropriate NHS Ethics committee. Participants took part in face-to-face interviews in both waves, administered using computer-assisted direct data entry, usually conducted in their own homes through the medium of English or Welsh with trained interviewers. The interview collects detailed information on health and disease, lifestyles, cognitive function, social networks, mental health, well-being and resilience, and demographics (for more details, see http://cfaswales.bangor.ac.uk/publications.php. en). Individual responses range from 1 (strongly disagree) to 7 (strongly agree). These are summed for a final scale ranging between 1 (low satisfaction/extremely dissatisfied) to 35 (highly satisfied), with high well-being defined as a score >26.

Depression and anxiety (absence indicates positive outcome)
These were each defined using the Geriatric Mental State Automated Geriatric Examination for Computer-Assisted Taxonomy (GMS-AGECAT) algorithm, 23 where a score of 2 indicated mild symptoms and a score of 3 or above indicated a case-level anxiety or depression. These were each dichotomised as 0 = no symptoms; 1 = symptoms present (mild and case level).

Social engagement
Social network resources, in the form of the size, closeness and frequency of contact with friends and relatives were measured by the six-item Lubben Social Network Scale. 24 Scores range from 0 (completely isolated/few social resources) to 30 (low isolation/ many social resources).
Based on Fortuijn et al. 25 breadth of social participation in group activities was ascertained through a range of questions providing an index ranging from 0 (no participation) to 6 (high levels of participation). These reflected activity across six domains (sports, political involvement, environmental groups, education or learning, arts, and voluntary or community groups).

Psychological factors
Abbreviated versions of three psychological factors are available in the CFAS Wales survey. Personal competence was assessed with six items derived from the Resilience Scale, 26 with responses ranging from 1 = strongly disagree to 5 = strongly agree. Self-esteem was assessed with eight items derived from the Rosenberg Self Esteem Scale (RSES), 27 with responses ranging from 1 = strongly disagree to 5 = strongly agree. Interpersonal control was derived from the Spheres of Control (SOC) scales 28 consisting of five questions, with responses ranging from 1 = strongly disagree to 5 = strongly agree. These items were selected based on psychometric 1022evaluation in a previous large cohort study of older people. 29 Subjective memory complaints were identified from the following questions 'Have you ever had any difficulty with your memory?' and 'Have you tended to forget things recently?' Following the rationale of Yates et al. 19 a positive answer to either question indicated a memory complaint, which was recoded into a dichotomous outcome (1 = no, 0 = yes).

Healthy lifestyle variables
A range of healthy lifestyle variables were derived in the previous analysis of the CFAS Wales data examining modifiable lifestyle factors and cognitive function. 17 Level of physical activity was determined by the reported frequency of engagement in 18 types of mild (e.g., light gardening, bowls, light housework), moderate (e.g., gardening, walking at a moderate pace, floor or stretching exercises), and vigorous (e.g., jogging, swimming, cycling) physical activity. A continuous scale was generated using the frequency levels (0 = once a year or less, 1 = several times a year, 2 = several times a month, 3 = several times a week, and 4 = every day or almost every day) multiplied by the intensity ratio (mild: moderate: vigorous = 1:2:3), based on the metabolic equivalent of task (MET) ratio. 17 Healthy diet represents the frequency of 'Mediterranean Style' food intake. Responses to each question ranged from never, seldom, once a week, 2-4 times a week, 5-6 times a week or daily.
A total score for healthy diet was generated based on the six levels of frequency. Alcohol consumption was determined with the question 'How often have you had an alcoholic drink of any kind in the last 12 months?' Responses ranged from 0 = not at all to 7 = almost every day.

| Outcome variable (loneliness)
Loneliness was assessed with the six-item De Jong Gierveld scale. 30 The scale ranges from 0 to 6 where a score of 0 to 1 indicates no loneliness, score of 2 to 4 moderate loneliness and 5 to 6 severe loneliness.
The scale has two sub-scales which measure emotional loneliness (the absence of an intimate relationship, such as a partner, best friend), and social loneliness (the absence of a broader social network such as siblings, cousins, friends and neighbours).  Table 1). To ascertain whether the missing data for subjective memory complaints' influenced estimations, the analysis was repeated without this variable.    Table 1 presents the descriptive statistics and results of the univariate analyses. It shows that males have better odds for MHR than females, whilst the odds of MHR were lower for those not married compared to those who were married. Of the psychological factors, the odds for MHR were greater for higher levels of self-esteem, interpersonal control and personal competence.

| Univariate predictors mental health resilience
Higher levels of social network resources were associated with greater odds for MHR, as were no subjective memory complaints.
In terms of healthy lifestyle, MHR was related to more physical activity.

| Multivariate analyses
Of the significant univariate predictors, the cumulative effects model ( Controlling for these significant determinants and wave 1 loneliness scores, MHR was significantly associated with lower total and sub-scale scores for loneliness at wave 2 (see Tables 2 and 3). Sensitivity analysis shows these effects held at lower levels of cognitive function when the MMSE was <25, but not at <23. The total score for loneliness was lower for those with more social resources and higher self-esteem at all levels of cognitive function.
Higher wave 2 MMSE scores were associated with lower loneliness total scores for the sample as a whole, but not in the sensitivity analysis. The effect of MHR on the total loneliness score was stronger than the total loneliness score at wave 1. In relation to the sub-scales (Table 3), emotional loneliness was lower at all levels of cognitive function for those with higher self-esteem and social loneliness was lower at all levels of cognitive function for those with more social resources.

| DISCUSSION
This study offers new insights using a measurement approach closely tied to contemporary conceptual understandings of resilience. Across two waves of data, just under a quarter of older people living with cognitive impairment sustained good mental health (no depression, no anxiety and high well-being), which we define as MHR. To our knowledge, this is the first exploration of MHR in this population and so is an important first step towards advancing theory and measurement approaches. Whilst a substantial body of research has examined resilience in younger populations, in comparison, the study of resilience in later life has not received the same attention. 11 We sought to identify the personal characteristics and wider social aspects important for resilience, 'the resilience reserve', through logistic regression. The sample size was sufficient in relation to 'rule of thumb' suggestions for logistic regression. 31 Here, we did not find strong statistical evidence that age and years of education were associated with resilience, but male and married/ cohabiting respondents were more likely to be resilient.
Potentially modifiable psycho-social factors were identified in the univariate analysis of the 'resilience reserve' that could be tar- Absence of subjective memory complaints predicted resilience in both univariate and cumulative effects analyses, suggesting that those who focus less on memory problems, perhaps appearing less aware of difficulties, also report better well-being and mood.
Although this effect may be interpreted as a form of positive response bias, it may also be viewed as an adaptive form of coping in some situations, focussing on strengths rather than problems. 33   with careers and family members. 39 We chose the complete case approach to reflect previous analyses using the CFAS Wales data. 3,40 Although 5.2% (n = 30) of participants were removed, the analysis was still sufficiently powered, but we recognise that undertaking a complete case analysis is not without criticism and is a limitation of the study. a pressing concern and could inform effective strategies for the public, health and social policy.