Health status risk factors and quality of life in 75–84-year-old individuals assessed for dementia using the short 10/66 dementia diagnostic schedule

Recruitment

The study used the validated 10/66 door-to-door interview protocols. Of the survey responders, we identified all individuals in our database (836 persons) who were considered middle-old in the selected electoral enumeration districts (ED). Diagnosis of dementia was derived from the Community Screening Instrument for Dementia (CSI-D; Hall et al., 1993; Hall et al., 2000) administered to participants and informants, the CERAD 10-word list learning task (Morris et al., 1989) and the EURO-D depression screen (Castro-Costa et al., 2008). Output from these instruments was then used to assign dementia status according to the short 10/66 algorithm (Stewart, Guerchet & Prince, 2016) which draws on the output of these instruments and applies a series of regression coefficients to assign a probabilistic diagnosis of dementia (Prince et al., 2003). In the short schedule, the Geriatric Mental State interview is replaced by the EURO-D scale. The algorithm produced output in 811 participants (97%), while the remaining 25 had missing source data that precluded algorithm output. A summary of the sampling methodology is shown in the flowchart below. Figure 1.

Measurements

In addition to their age and gender, participants were asked to report on their ethnicity, marital status, highest level of education, religion, occupation, type and tenure of accommodation, and living arrangements (See File S1). As part of the survey an informant questionnaire was also done. Ethnicity was recorded according to the most common groups present within the island, including African, Caucasian, Chinese, East Indian, Mixed African and Indian, Indigenous, Mixed Other, Portuguese, Syrian/Lebanese, and others. These were re-coded for analyses as African, East Indian, Mixed African and East Indian, and Mixed other/Others. Religion was recorded as 15 separate denominations including categories for “None” and “Not Stated.” These were then coded as Christian, Hindu, Muslim and Others and “Not Stated” as a group comprising other religions for analyses. Marital Status was categorised as married/co-habiting, divorced/legally separated, never married and widowed. Education was coded as none, primary (i.e., formal education ages 5 to 11 years), secondary (i.e., formal education from 12 to 19 years), or tertiary and other (i.e., education at the university or polytechnic level). Reported previous occupation was categorised as agricultural worker, associate professional, clerical worker/secretary, housewife, manager/administrator, professional, semi-skilled labourer, skilled labourer or unskilled labourer. Previous occupation referred to the major occupation held during working years. These categories were then merged for analyses as associates/professionals/managers, agricultural workers, unskilled or semi-skilled labourers, skilled labourers, and housewives. Type of accommodation were categorised as detached house, and condominium, townhouse, or other. Tenure of accommodation was categorised as privately owned (including by spouse or family), and rented (including from private landlord, National Housing Authority (NHA), housing), or other. Living arrangements of chief respondent were categorised as self (i.e., living alone), with spouse, with spouse and children, with children alone, and with other.

Questions were also asked about the frequency of smoking and alcohol consumption. Participants were asked if they had any of the following 15 medical conditions listed in their lifetime and which persisted in the last 3 months: angina, Alzheimer’s disease, arthritis, asthma, cancer, dementia, depression, diabetes, epilepsy, head injury, heart disease, high cholesterol, hypertension, Parkinson’s disease and stroke. These were then categorised as presence of number of co-morbidities in a person which were coded as none, 1, 2, 3 and greater than/equal to 4. Self-reported medical conditions were recorded as yes/no for arthritis, angina, diabetes, heart disease, high cholesterol, hypertension/BP medicine and stroke in the analyses. The following impairments were also recorded as binary responses: eyesight, hearing, climbing stairs, walking, and ability to speak and understand. An IADL score was generated based upon binary yes/no responses on ability to perform the following tasks: using the telephone, shopping independently, preparing food independently, doing housekeeping, doing laundry, commuting independently, handling finance independently and taking own medicine (Lawton & Brody, 1969). The IADL score ranged from 8 (able to perform all the activities) to 0 (cannot perform any of the activities) (Juva, Makela et al., 1997). Participants were also asked if they considered themselves to be physically active. Physical activity being defined as any bodily movement that is produced by the contraction of skeletal muscle and that substantially increases energy expenditure (CDC, 2017). The responses were recorded as not at all, not very much, fairly, and very much. These were re-coded as no exercise, low exercise, and moderate to high level of exercise for analyses.

Statistical analyses

Prevalence of dementia was calculated and reported. Univariable and multivariable analyses of factors associated with dementia were carried out using logistic regression on the unweighted sample. To address the non-independence of EDs selected, robust clustered standard errors were used. All analyses are reported in tables with odds ratio (OR), 95% confidence intervals (95% CI), robust standard errors (SE) and p-values. The level of significance used was two-sided p < 0.05. Stata 14.2 (StataCorp, 2015) was used for all analyses.

We applied univariable logistic regression to model the association between dementia and self-reported medical conditions and quality of life variables. A series of models were developed using logistic regression methods. The models were assessed for calibration and discrimination using area under the Receiver Operating Characteristic (AUROC), goodness-of-fit (GOF) and Hosmer and Lemeshow’s tests. Calibration is a measure of how well predicted probabilities agree with actual observed risks while discrimination is a measure of how well the model can separate those who do and do not have the disease of interest (Cook, 2007). The larger the AUROC curve and the closer to 1, the better the diagnostic effect of the model (Kleinbaum & Klein, 2010). We fitted multivariable logistic regression on self-reported medical conditions related to vascular diseases, arthritis and diabetes (Model I). To address non-linear associations, age squared was added to the model. Age is a known risk factor for dementia, while gender, ethnicity and religion may be confounders for angina, diabetes, heart disease, high cholesterol, hypertension and stroke patients since the incidence, time and length of disease may be effected by hormone levels, ethnic make-up and dietary practices. Hence, the model was adjusted for age, age², gender, ethnicity and religion. Education and occupation were then added to the model to account for differences in lifestyles arising out of educational attainments and practising occupation for the majority of their lives in Model II. In the third model, we included age, age², gender, ethnicity, religion, education, occupation, marital status, living arrangements, type of accommodation, smoking and alcohol consumption to assess their association with dementia (Model III). These additional factors were added to account for their loneliness, ability to look after themselves, avoid smoking and limit their alcohol consumption to acceptable levels. A similar approach was taken for developing multivariable models for quality of life variables (impairments, IADLs and exercise) adjusting for age, age², gender, ethnicity and religion (Model IV), adjusting for age, age², gender, ethnicity, religion, education and occupation (Model V), and adjusting for age, age², gender, ethnicity, religion, education, occupation, marital status, living arrangements, type of accommodation, smoking and alcohol consumption (Model VI). The best models were chosen based model specification tests and performance.

Univariable analyses

Age, gender, ethnicity and religion are non-modifiable risk factors for dementia and multiple morbidities. Our univariable analyses confirmed that women, those of East Indian descent and followers of Hindu faith were more likely to suffer from dementia. Among the middle-old population, dementia was associated with a lower level of education. The majority of individuals, about seventy five percent in the studied age range born between 1936 and 1945, had no educational qualifications. This picture of dementia in the elderly may change in the future reflecting mandatory schooling up to the secondary level in the younger generation in Trinidad. Education is a potentially modifiable risk factor that is believed to build cognitive reserves. Occupation was also a factor associated with dementia status. Being an agricultural worker, skilled labourer or housewife puts an individual at a higher risk for having dementia. Challenging jobs may decrease rates of cognitive decline and also build up an individual’s cognitive reserve. The findings of our study are consistent with reports that jobs that are both challenging and afford opportunities to take responsibility are associated with a reduced likelihood of dementia (Then et al., 2015; Qiu & Fratiglioni, 2018), although the explanation might also lie in better access to healthcare and the adoption of healthier lifestyles by those in higher educational and income categories. Individuals with more years of education might also be more likely to access more challenging jobs.

Alcohol consumption compared to non-consumption was associated with lower odds of dementia, consistent with other research reporting that alcohol consumption is associated with a lower risk of any dementia or vascular dementia (Letenneur, 2004). Conversely, heavy alcohol binge drinking has been reported to be associated with increased risk of dementia (Järvenpää et al., 2005). In Trinidad and Tobago society, many consider it is a social norm when meeting with friends and family to have a few alcoholic beverages (Sinanan, 2017). The findings with respect to alcohol may, therefore, reflect the effect of greater social interaction. Those living with children alone were more likely to have dementia than those living alone by themselves, although this might reflect a higher level of dependency in the former group accounting for their living arrangements.

Those individuals having more than four co-morbidities were more likely to have dementia than those with no co-morbidities. Multimorbidity increases with age, as does dementia; such individuals are admitted to and stay longer in hospital with the attendant increase in complications, cost and mortality as well as increased use of health care services. Multimorbidity can reduce the quality of life of older people, and when dementia is a part of the condition, the situation may be even more demanding. Our univariable analyses showed that those with self-reported diabetes and stroke were more likely to suffer from dementia. This agrees with a study that found that multimorbidity is associated with increases in functional limitations, and the associations are stronger among women than among men and among adults aged 75 or older than among those aged 65 to 74 (Jindai et al., 2005). Age and gender will always play a role as will the duration of each co-morbidity because chronic diseases have time-varying effects.

Further, our univariable analyses revealed that impairment in eyesight, hearing, climbing stairs, walking and ability to speak and understand were more common in those having dementia. Individuals with dementia were more likely to be unable to perform all eight IADLs. Those who did not exercise at all and, by extension, lived sedentary lifestyles were also more likely to have dementia. Of the participants, 24% were assessed by the algorithm as having dementia and more than half of these individuals had impairment in IADLs. Shopping and commuting (Fig. 2) were among those activities that were adversely affected in participants with dementia. Studies have shown that there are differential effects of the type and frequency of social participation on IADL decline in older people (Tomioka, Kurumatani & Saeki, 2018) and that the type and frequency of social participation associated with IADL decline varies according to physical and mental functioning (Tomioka, Kurumatani & Saeki, 2018; Griffith et al., 2017).

Multivariable analyses

We developed a series of multivariable models to identify the self-reported medical conditions and quality of life variables associated with dementia. Our multivariable analyses of self-reported medical conditions adjusting for age, age², gender, ethnicity, religion, education and occupation (Model II) showed that those with self-reported diabetes and stroke were more likely to have dementia than those who did not have diabetes and stroke, respectively. This is not surprising in a country with high vascular risk, and is comparable to findings from another study (Chadee et al., 2013); however, it should be noted that their study population was limited to those who accessed a short-term financial assistance programme run by the Ministry of the People and Social Development in the Government of Trinidad and Tobago. The risk of dementia has been also associated with the duration and severity of diabetes (Livingston et al., 2020). Qiu & Fratiglioni (2018) suggest that interventions targeting multiple major vascular risk factors may represent a promising approach in reducing dementia risk and delaying its onset. It is also noteworthy that in our study only 14% of individuals with no comorbidity had dementia which was similar to that reported by another study (Browne et al., 2017).

Our multivariable analyses of quality-of-life variables adjusting for age, age², gender, ethnicity, religion, education and occupation (Model V) revealed a persisting association of impairments with dementia. Qiu & Fratiglioni (2018) report that people who frequently engage in mentally stimulating activities (e.g., learning, reading, or playing games) from young adulthood through midlife and old age are less likely to develop dementia. It is entirely possible that physical impairment impacted negatively on the ability of an individual to be active. Focusing on healthcare activities that may lead to reducing the likelihood of impairment such as eye care in diabetes and encouraging more activity would be beneficial in our middle-old population. Some studies have reported that aerobic exercises are associated with reduced risk of cognitive impairment and dementia, (Matura et al., 2017) and that physical activity not only improves fitness but protects brain cells (Ahlskog et al., 2011). However, patterns of physical activity change with age, generation, and morbidity and are different across sex, social class, and cultures. Meta-analyses of longitudinal observational studies of 1–21 years duration showed exercise to be associated with reduced risk of dementia (Livingston et al., 2020). The associations between leisure activity, not smoking, and increased survival still existed in those aged 75 years or more, with women living longer than men (Rizzuto et al., 2012). This is an important piece of information for caregivers and policymakers to take into consideration when planning care strategies.

Strengths and limitations

Strengths of the study included its large size (allowing sufficiently precise estimates, as indicated by the widths of the confidence intervals) and household enumeration methodology to achieve a representative sample of the source population. A standard and widely used protocol was also used for dementia assessment and ascertainment. Considering potential limitations, this study was based upon face to face door-to-door interviews with individuals willing to participate. This may have introduced a selection bias at the participation stage, as well as through further missing data precluding dementia algorithm output (although the latter only accounted for 3% of participants lost). Although corroboration was sought from informants on exposures where possible, there may have been error arising from self-reporting bias, although this would more likely add error and obscure rather than exaggerate associations of interest. The data presented here were drawn from a larger study whose aim was focused on measuring the prevalence of dementia in Trinidad; thus not all measures of potential interest were available. Gender could also be treated as an effect modifier for dementia and multiple morbidities. The data also did not allow us to explore the effect of the environment on IADL, or dietary or other lifestyles as risk factors (Hendrie et al., 2001; Smith, Petot & Perry, 1999), or the relationship between APOE e4 and education or occupation. The survey did not collect information either on the duration of risk factors or the onset of dementia.