Does Aboriginal ancestry predict cognitive performance scale scores? A multilevel analysis of resident assessment instrument for home care data

Abstract

This study examined the relationship between Aboriginal ancestry and Cognitive Performance Scale (CPS) scores utilizing data from the Resident Assessment Instrument for Home Care (RAI-HC). After controlling for demographic variables, individual ancestry was not predictive of cognitive status. However, CCACs servicing a higher proportion of Aboriginal clients had lower CPS scores. Future research should focus on the individual and environmental factors which affect cognitive status in Aboriginal peoples.

Introduction

Little is known about the epidemiology of Alzheimer’s Disease (AD) and non-Alzheimer’s dementia (NAD) in the Aboriginal population. The few studies that exist on this topic suggest that Aboriginal peoples have a lower incidence and prevalence of AD. For example Hendrie and colleagues (1993) found the age-adjusted prevalence of AD was 0.5% in an on-reserve First Nations sample, significantly lower than 3.5% observed in the control group. When other dementias were included (e.g., multi-infarct dementia) the prevalence was 4.2% for both groups (Hendrie et al., 1993). In a sample of 26 age-matched Cherokee peoples with and without AD the prevalence of AD decreased as genetic Cherokee ancestry increased, suggesting an age-related protective effect for Cherokee peoples (Rosenberg et al., 1996). A study that examined the course of AD in American Indians, Caucasian, and African-American people found no differences in age of onset and years from symptom onset to death (Weiner et al., 2007). There have been no evaluations of treatment outcomes specifically involving Aboriginal peoples. As a whole, research on dementia and Aboriginal peoples is in its early stages (Jervis & Manson, 2002).

One difficulty in establishing relationships between ancestry and dementia is that ancestry is frequently confounded with another variable associated with dementia risk: socio-economic status (SES; Griffith & Griffith, 2008). Although SES has been measured in a variety of ways, such as education, income, and occupation, it has consistently predicted AD and NAD (c.f. Katzman, 1993; McDowell, Xi, Lindsay, & Tierney, 2007). In particular, low SES is associated with presence of AD and NAD. Thus, studies of dementia in Aboriginal populations must take into account this variable, as levels of education, employment and income are generally lower in the Aboriginal populations (Adelson, 2005; Tjepkema, 2002). The purpose of the present study was to examine the relationship between cognitive status and Aboriginal ancestry after controlling for social and clinical variables using data from the Resident Assessment Instrument for Home Care (RAI-HC).

Methods

Study Sample

A total of 133,286 adult clients from 43 Community Care Access Centres (CCACs) were assessed with the RAI-HC between April 1, 2004 and March 1, 2005. The CCAC’s are regional agencies that co-ordinate assessments and services for home care as well as long-term care, and reasons for such assessment include screening for home care eligibility and placement on wait lists for long-term care. The study sample (n = 126,381) included those clients with complete ancestry data. The RAI-HC is a multidimensional assessment and care planning tool that incorporates clients’ physical and psychological functioning (cf. Hawes et al., 1995; Morris et al., 1999; Morris et al., 1997). It has been translated into many languages and is currently used in Canada, the United States, Japan, China and many countries in Europe (Hawes et al., 1997; Sgadari et al., 1997).

Outcome Variable

The RAI-Cognitive Performance Scale (CPS) combines five items related to cognition to form a single scale with seven categories of cognitive impairment (0 = intact; 6 = very severe impairment). The five RAI-HC items consist of short-term memory, cognitive skills for daily decision making, making self understood, self-performance in eating, and comatose status. The reliability and validity of the CPS are considered good (c.f. Morris et al., 1994; Landi et al., 2000).

Predictor Variables

The predictor variable of interest was Aboriginal ancestry (i.e., Inuit, Métis, or North American Indian) and was coded as a dichotomous variable (“individual ancestry”). Another ancestry variable was created to represent the proportion of Aboriginal peoples serviced by the CCAC; this variable was termed “agency ancestry.” This contextual variable was created to determine if the proportion of Aboriginal people in a CCAC affected CPS scores. Age was entered as a continuous variable and the highest level of education attained was entered as a categorical variable (less than high school education, high school, or post-secondary education). Additional dichotomous variables included the presence/absence of AD, NAD, stroke, head trauma, Parkinson’s disease, multiple sclerosis (MS) and depression.

Analyses

All data were analysed using SAS 9.1. Descriptive statistics were tabulated separately for Aboriginal and non-Aboriginal clients. The outcome variable of interest was client’s CPS scores. Due to the nested structure of the data (clients nested within CCACs), a multilevel regression analysis was used. The CCAC variable was considered a random group-level predictor. Fixed individual-level variables consisted of cultural background (ancestry), demographic (age, sex, education) and clinical (AD, NAD, stroke, head trauma, Parkinsonism, depression) variables. Ancestry was introduced as an individual characteristic (i.e., “individual ancestry”) and also as a contextual characteristic (i.e., “agency ancestry”).

Multilevel linear modeling using SAS mixed analysis version 9.1 was used to build sequential models for the dependent variable. First, a null model was analyzed to determine if a difference in the dependent variables existed between CCACs. Next, individual ancestry was entered as a predictor to examine its effect on the dependent variable. Additional control variables were entered in ascending complexity until the model failed to converge. Failure to converge indicated a poor fit between the model and the data, and variables that caused convergence failures were removed from the model.

Results

A total of 126,423 individuals from 43 regional agencies were included in the analyses. Table 1 displays demographic and clinical characteristics of the sample by ancestry. The overall prevalence of AD was 6.68%, with AD being monitored/treated in 0.96% of clients and present but not monitored/treated in 5.72% of clients. The overall prevalence of NAD was 9.43%, with NAD monitored/treated in 1.34% of clients and present but not monitored/treated in 8.09% of clients.

Multilevel linear models were built sequentially (as described above) to predict clients’ CPS scores. Table 2 displays the statistical results for each model that converged. In the null model (Table 2 model 1) CPS scores varied with CCACs and so CCAC was included as a random variable throughout the model-building process. The next model included individual ancestry as a random and fixed variable; this model did not converge (results not shown). In the next model (Table 2, model 2) individual ancestry was predictive of CPS scores; Aboriginal clients had lower CPS scores than non-Aboriginal clients. A third model (Table 2, model 3) added agency ancestry as a fixed predictor; this variable did not predict of CPS scores.

Next, demographic variables were added to the model as controls. Once these control variables were added to the model, individual ancestry was no longer predictive of CPS scores (Table 2, model 4). Clients of younger age, male sex, and lower educational attainment had higher CPS scores. In the final model, diagnosis of AD, NAD, head trauma, stroke, Parkinsonism, MS, and depression were added as predictor variables (Table 2, model 5). Presence of any of these diagnoses was associated with higher CPS scores.

Interestingly, when these clinical variables were included in the model, agency ancestry became predictive of CPS scores; the CCACs with proportionally fewer Aboriginal clients had higher CPS scores. Lower educational attainment and male sex were associated with higher CPS scores, while age was not.

Discussion

This study examined the relationship between ancestry and cognitive status scores in a population of people assessed for home care and long-term care placement. Of particular interest was whether Aboriginal ancestry was predictive of poor CPS scores after controlling for key demographic variables such as SES. Preliminary models using data from the present study indicated that Aboriginal clients had lower CPS scores, indicating better cognitive status when compared to non-Aboriginal clients.

However, after controlling for key demographic predictors of cognitive status such as age, sex and education, individual ancestry was no longer predictive of CPS scores. That is, cognitive status was similar between the two ancestry groups once demographic variables were controlled for. In the final model clients of male sex and lower education had higher CPS scores, indicating poorer cognitive status. Low SES has been linked to poorer cognitive status elsewhere in the literature (e.g., Katzman, 1993; McDowell et al., 2007).

Although individual ancestry was not predictive of CPS scores, when it was used as a contextual variable the proportion of Aboriginal clients serviced by the CCAC predicted CPS scores. More specifically, CCACs that serviced smaller proportions of Aboriginal clients had higher CPS scores. The interpretation of this finding is unclear and may be an artefact of the different age structure of Aboriginal clients in this population (44% of Aboriginal clients were under 65 years of age, compared to 17% of non-Aboriginal clients). Thus the number of Aboriginal people accessing services at CCACs and who were old enough to be vulnerable to AD and NAD in later life was relatively small. This difference could explain why CCACs that serviced more Aboriginal clients had lower CPS scores compared to those that serviced fewer Aboriginal clients. Future research should examine this contextual variable in greater detail.

Previous research suggested that Aboriginal peoples have a lower prevalence of AD (Hendrie et al., 1993; Rosenberg et al., 1996). Although the current study did not predict AD directly, it did find similar cognitive status scores for the Aboriginal and non-Aboriginal groups. If CPS scores were to be used broadly as a proxy for dementia (i.e., AD and NAD) these findings suggest that Aboriginal and non-Aboriginal clients had a similar prevalence of dementia.

Future research should examine the relationship between ancestry and dementia more specifically as a lower incidence of AD and higher incidence of NAD in the Aboriginal population may result in these same findings. Thus, the question of whether Aboriginal clients had similar or different prevalence of AD and NAD remains unanswered.

The few studies investigating dementia in Aboriginal peoples utilized people living on reserves; relatively little is known about the cognitive status of the off-reserve population. The present study adds to our knowledge about off-reserve Aboriginal people, as CCACs only provide assessment and screening services to Aboriginal people living off reserve. This study also stands out as population-based study; the provincial data was obtained for all people assessed with the RAI-HC within a one-year period. Previous research was limited by small samples of Aboriginal people, or a limited geographic area (Hendrie et al., 1993; Rosenberg et al., 1996).

Finally, the multilevel linear modeling approach used in the present study allowed for examination of individual- and CCAC-level predictors of cognitive status. This approach is ideal for analysis of RAI data as it capitalizes on the hierarchical structure of the data.

There are several caveats to the present study. The first caveat is with respect to classification of clients as “Aboriginal” and “non-Aboriginal.” Other researchers have commented on the significant variations within the Aboriginal population (c.f. Kirmayer et al., 2003; Waldram et al., 2006) which makes the clustering of all Aboriginal people into one category problematic. A similar argument can be made for the non-Aboriginal population. As the data are cross-sectional in nature it limits causal inferences. Similarly, the data are generalizable to the client population accessing services through CCACs and thus are not representative of the population as a whole.

Griffen-Pierce and colleagues (2008) provided an excellent overview of the challenges in assessment and diagnosis of AD in Aboriginal peoples, many of which are relevant to the current study. Cultural understandings and interpretations of cognitive decline may hinder the recognition, assessment and diagnosis of AD. Culture-specific issues in mental status testing pose further problems (Ferraro, 2001; Griffen-Pierce et al., 2008). For example, observations from interviews using a culturally adapted version of the Mental Status Questionnaire found that questions measuring awareness of place and time were not accurate indices of mental status as things such as postal address and calendar time were not significant to elderly First Nations people living in remote northern communities (Kaufert & Shapiro, 1996). Linguistic, cultural, and contextual factors in determination of mental status may have affected the RAI data for Aboriginal clients. The CPS has not been validated for use with the Aboriginal population and thus its utility with this population is unknown.

The present study demonstrated the importance of controlling for SES when examining the relationship between ancestry and cognitive status. As described by Griffith and Griffith (2008) perhaps the greatest utility of ancestry as a category is to capture the environmental factors that affect group members’ health status. Aboriginal ancestry as an individual variable was not predictive of cognitive status after controlling for demographic variables. However, Aboriginal ancestry as a contextual variable was predictive of cognitive status; CCACs with higher proportions of Aboriginal clients had lower CPS scores. Future research should focus on the individual and environmental factors which affect cognitive status in Aboriginal peoples.