What is the comparative health status and associated risk factors for the Métis? A population-based study in Manitoba, Canada

Study Population and Data

The health status and disease burden of the Métis in Manitoba were compared to all other residents of the province of Manitoba, Canada using a population-based approach. MCHP houses sets of data collectively referred to as the Population Health Services Data Repository. This Repository is an extensive, person-level, linkable yet de-identified Repository of administrative databases for the entire population of Manitoba, covering both health and social services records. Since the provincial Ministry of Health provides comprehensive universal health care coverage for essentially all residents of Manitoba (approximately 1.2 million people), non-participation in the plan is rare and claims data are nearly complete for the population [19, 22].

The Repository contains information on mortality and birth, physician and hospital use, pharmaceutical use, use of services such a home care and nursing homes (i.e., long term care facilities) and information derived from education and social services. All of this information is linkable by a scrambled and fictitious unique personal health information number (PHIN). As well, enumeration area information from Canadian Census data, such as neighbourhood-level average household income, can be attributed to the population at an aggregate level via the residential six-digit postal code. This study received permissions from the Faculty of Medicine's Health Research Ethics Board, and from the Health Information Privacy Committee of the Government of Manitoba, as well as from MMF for use of the Métis cohort database.

For mortality indicators including total mortality, premature mortality (death before the age of 75), mortality due to injury, life expectancy, potential years of life lost and all-cause mortality for individuals with diabetes, deaths were measured over calendar years 2002-2006 using Vital Statistics data, which includes cause of death. Ten years of data, 1997-2006, were required for deaths due to injury to obtain a stable rate.

Statistical Methodology

Age- and sex-adjusted rates of mortality and prevalence of illness were calculated using a generalized linear model (GLM) framework for each indicator. Data were analyzed by both the provincial Regional Health Authority (RHA) boundaries, to serve the needs of provincial health planning, and by the seven MMF Regions, to serve MMF socioeconomic program planning (see RHAs and MMF Regions in Figure 1). The adjusted relative risk of death or illness for the Métis in Manitoba compared to all other residents of the province was estimated in Poisson or negative binomial regression models adjusting for age and sex. A variety of age groups or linear and quadratic age terms were employed for various models to ensure the best model fit. Adjusted rates were then obtained by multiplying the estimated relative risk by the appropriate reference rates. Two exceptions are: life expectancy for males and females, which was calculated using a life table methodology based on deaths in the Manitoba population over five calendar years (2002-2006); and potential years of life lost, which was age- and sex-adjusted using direct standardization rather than a GLM, due to the age dependency of the outcome.

For selected indicators, further analyses included explanatory variables beyond age and sex. Logistic regression models were performed for the selected indicators of diabetes prevalence and lower limb amputation rates among people with diabetes, to further explore: (1) differences between Métis and all other residents; and (2) differences among the Métis only. Additional covariates included in the logistic models were: geographical location of residence, average household income of the neighbourhood based on the 2001 Census, major physical illness comorbidity and mental illness comorbidity, and continuity of care. For the model of all Manitobans, aggregate geographical areas were used - the two urban centres of Winnipeg and Brandon, plus an aggregate of North (Nor-Man, Burntwood and Churchill RHAs), Mid (Interlake, Parkland and North Eastman RHAs), and Rural South (Assiniboine, Central and South Eastman RHAs). The model of Métis only used the seven MMF Regions. Comorbidities were measured by the presence of one or more major physical or mental illness Aggregated Diagnostic Groups (ADG) calculated from Johns Hopkins' Adjusted Clinical Group (ACG) case mix system. Major physical ADGs included: ADG 3 = Time Limited: Major, ADG 4 = Time Limited Major-Primary Infections, ADG 9 = Likely to Recur: Progressive, ADG 11 = Chronic Medical: Unstable, ADG 16 = Chronic Specialty: Unstable-Orthopedic, ADD 22 = Injuries/Adverse Effects: Major and ADG 32 = Malignancy. Mental illness ADGs included: ADG 23 = Psychosocial: Time Limited, Minor, ADG 24 = Psychosocial: Recurrent or Persistent, Stable and ADG 25 = Psychosocial: Recurrent or Persistent, Unstable. Continuity of care was measured as the proportion of residents with greater than fifty percent of their ambulatory care visits to the same physician over a two year period. In the logistic regression models, we have 80% power to detect an OR (odds ratio) of 1.038 (or its inverse 0.963) for the diabetes analysis, and 1.279 (or its inverse 0.744) for the amputation due to diabetes analysis. All models had acceptable goodness of fit tests, with a c-statistic ranging from 0.743 to 0.773. All analyses were performed on a Unix server with SAS version 9.1 (SAS Institute Inc, Cary, North Carolina).

Mortality and morbidity

Life expectancy is not as sensitive an indicator, and is only 2% lower for Métis males (75.0 versus 76.8 years, p < 0.001), but not statistically significantly different for females (81.0 vs. 81.8 years, p = 0.061; NS). PYLL is 22% higher (64.6 vs. 53.1 years per 1000, p < .0001) for the Métis, as is premature mortality rate (4.0 vs. 3.3 deaths per 1000 before age 75, p < 0.001).

Morbidity is also elevated for the Métis compared to all other Manitobans, with the exception of osteoporosis (see Table 3). The highest disparity occurs for diabetes and related complications - diabetes prevalence is 33% higher for the Métis (11.8% vs. 8.8%, p < 0.001), and lower limb amputation due to diabetes is 49% higher (24.1 vs. 16.2%, p < 0.001). Not surprisingly, kidney dialysis is also elevated, at 35% higher (0.46% vs. 0.34%, p < 0.001). Related cardiovascular diseases, including acute myocardial infarction (AMI) (5.4 vs. 4.3 per 1000, p < 0.001) and stroke (3.6 vs. 2.9 per 1000, p < 0.001) are also more common among the Métis. Manitobans with diabetes have elevated all-cause five-year age- and sex-adjusted mortality rates, but this is even higher in the Métis population compared to all other Manitobans (20.8% vs. 18.6%, p < 0.02). Of note, if the Métis were compared to only the 'all other Manitoba population that do not have aboriginal ancestry, one might expect an even wider mortality and morbidity gap than reported in this study. For example, the poorer health status and reported four-fold differences in diabetes prevalence in First Nations compared to all other Manitobans [24, 25] may have been a major contributing factor in the 'all other Manitoba' prevalence of diabetes in this study.

Regression models of diabetes and lower limb amputation

The regression modeling of the likelihood of having diabetes (see Table 4, All Manitobans section) shows that the elevated diabetes prevalence of the Métis persists (aOR 1.29, 95% CI 1.25-1.34, p < .001) even when controlling for the effects of geography, age, sex, neighbourhood income, and other mental and physical comorbidities. Looking at the predictors of diabetes within the Métis population alone (Table 4, Métis only section), the likelihood of diabetes increases with increasing age (aOR 1.20, p < .001), with a plateauing effect at the older age category, as indicated by the quadratic age variable being below 1. Interestingly, there is no statistically significant difference in prevalence by sex, or by mental illness comorbidity. Neighbourhood income, meaning the average household income of the census enumeration area in which the person resides, shows an inverse relationship, with the higher income areas having lower diabetes prevalence (aOR 0.88, p < .001). There is a high degree of association with comorbid physical illnesses (aOR 1.65, p < .001). The effect of geography is quite profound, with Métis living in the northern MMF regions having the highest likelihood of being diagnosed with diabetes (The Pas Region aOR 1.22, p < .001; Thompson Region aOR 1.66, p < .001). All other MMF regions have rates statistically significantly lower than the overall Métis provincial average. This mirrors the information in Table 4 for all Manitobans, where the North aggregate area shows elevated diabetes prevalence for the entire population of Manitoba, both Métis and others.

Table 5 (all Manitobans section) includes the analysis of the likelihood of one particular adverse outcome of diabetes - lower limb amputation. Comparing Métis to all other Manitobans, there is no statistically significant difference (aOR = 1.13, 95% CI 0.90-1.40, p = 0.29) in the likelihood of Métis having an amputation, once controlled for other effects (geography, age, sex, income, continuity of care, and mental/physical comorbidities). An increased likelihood of amputation is associated with older age (with a plateauing effect at older ages), being male (aOR = 1.94, p < .001), and having major comorbid physical illnesses (aOR 3.25, p < .001), but mental illness comorbidities are not statistically significantly associated. As well, residing in a higher income area, and having continuity of physician care both decrease the likelihood of having a lower limb amputation due to diabetes. When the analysis is limited to Métis only (see Table 5, Métis only section), similar associations are seen, with age, being male and having major physical comorbidities increasing the likelihood of amputation, and the two factors of residing in a higher income area and having continuity of physician care decreasing the likelihood.

Although the likelihood of amputation is statistically significantly higher in certain geographical regions (i.e., living in the Mid or North area of Manitoba) as shown in Table 5 for all Manitobans, this same table shows no statistically significant effects of geography for the Métis only. That being said, a similar trend of elevated likelihood of amputation in the two northern MMF regions (The Pas and Thompson) may indicate type 2 error, i.e., a small sample size not having enough power to detect a statistically significant difference.

Limitations

One limitation of this study is reliance on the use of administrative claims data alone. Although there have been validation studies completed previously [23], there may potentially be undercounting of diagnoses for those people not seeking medical help. However, that would not be any more likely for those people living in similar geographical locations, so the differences in health status should not be affected. The Canadian universal health care system also equally applies to all, so lack of access through potential income barriers would not be as problematic as in a country without universal health care provisions. A further limitation may be in the establishment of the Métis cohort used in this study. This may have included a small number of people who are not Métis, but who were included through familial relationships using health registry data available to us (for example, if a non-Métis married a Métis, we would have classified both as Métis). If anything, this would most likely reduce the gaps in health status between Métis and all other Manitobans, so that the health inequity found in this study would be maintained or increased if a more accurate identifier were available. That being said, the Métis population count obtained through our method was similar (i.e., less than 2% difference) to that obtained through self-report of the Census.

Age- and sex-adjusted morbidity and mortality rates

Mortality and morbidity rates are, in general, higher in the Métis population of Manitoba compared to all other Manitobans when we look at age- and sex-adjusted rates. Mortality rates, whether premature mortality, total mortality, injury mortality, or potential years of life lost, all appear to be more sensitive indicators of differences than life expectancy. The former mortality indicators show elevated rates for Métis of 14% to 23%, whereas life expectancy is not significantly different for female Métis, and only 2% lower for male Métis compared to the rest of the population. In terms of morbidity, diabetes (34% higher for Métis) and ischemic heart disease (40% higher) show the biggest gap in a disease outcome for adults aged 19 and older. As well, age- and sex-adjusted diabetes-related outcomes such as lower limb amputations with diabetes comorbidity (49% higher), and dialysis initiation (35% higher) all mirror the elevated diabetes rate for the Métis.

Our research shows higher Métis life expectancies than those reported previously [9–11], and a gap of less than 2 years difference between Métis and all other Manitobans. As well, there appears to be a much smaller gap than found for Manitoba First Nations people, where life expectancy was 8 years lower than the rest of the population [24, 25].

In our study, the overall Manitoba age- and sex-adjusted prevalence of diabetes was elevated for Métis compared to all other Manitobans (11.8% vs. 8.8%; Relative Risk [RR] = 1.34). Although the prevalence of diabetes in the Métis population is elevated in Manitoba, this report did not find a doubling or tripling effect as in other Métis studies previously [10, 13, 14, 16]. This may relate to the fact that there may be underlying undiagnosed diabetes, or the rest of the Manitoba population rate is much higher than in some other provinces, or the Manitoba sample in previous studies was different than our population-based cohort that included all Manitoba Métis people. For those living with diabetes, we found an elevated risk of lower limb amputation for the Métis compared to the rest of the population (24.1 vs. 16.2 per 1000) when adjusting only for age and sex. Although a gap exists, in fact, many of the health indicators for the Métis appeared to be somewhat between those rates found for the general population and for the First Nations populations [24–26].

A more complex analysis of diabetes and amputation - regression modeling to determine factors

The main predictors of diabetes for the Métis were similar to those found in previous research studies, with older age and physical comorbidities both being strongly associated with increased likelihood (see Table 4). The sex difference in the diabetes model was not statistically significant for Métis (aOR = 1.00, 95% CI 0.94-1.07, NS), and only slightly elevated for males in the entire population model (aOR = 1.14, 95% CI 1.12-1.16). This corresponds with the finding of Janz et al. [16], but is contradictory to the finding of Bruce where females had a higher prevalence [13, 14].

However, the sex difference for amputation due to diabetes showed a high degree of elevated risk for males for both the entire population model (aOR = 1.94, 95% CI 1.71-2.21) and the Métis only model (aOR = 2.36, 95% CI 1.50-3.71). Although this study did not look at causes of the sex differences, it could be speculated that males have a more rapid progression of disease or may be diagnosed with diabetes at a later stage, thus being more at risk for amputation. Altenberg et al. [27] have demonstrated that people exhibiting lower health-conscious behavior, visiting a healthcare provider less often or showing less anxiety about their diabetes are more likely to have diabetic foot ulcers (which, in turn, could lead to greater risk of amputation). So factors such as these may be more likely in males, hence they would have an elevated risk for amputation - this needs further study in Manitoba to understand the sex differences.

For the Métis only (see Table 4), in contrast to Bruce [13, 14], a lower likelihood of having diabetes was associated with higher average household income of the neighbourhood (aOR = 0.875, 95% CI 0.86-0.89 for each $10,000 increase). As well, Métis living in the southern and mid-provincial MMF Regions had less likelihood of diabetes, but Métis living in the two northern MMF regions (The Pas MMF Region aOR = 1.22, 95% CI 1.11-1.33; Thompson MMF Region aOR = 1.66, 95% CI 1.49-1.86) were more likely to have diabetes compared to the provincial average. Bruce et al. [17] did not find a geographical difference, but this may have been a type 2 error due to a much smaller sample size. That study did show a trend towards higher diabetes prevalence in rural/northern compared to urban areas (7.1% vs. 5.7%, NS), albeit not statistically significant. Once again, further study is required to understand why the likelihood of diabetes is higher for Métis living in the North. This is contrary to the finding of Martens et al. [24], where First Nations people had a greater risk of diabetes in the southern Tribal Councils, possibly due to a longer history of colonization, stress and lifestyle changes in more urbanized areas compared to more remote areas.

Comparing the age- and sex-adjusted models with the more complex models

The regression models of diabetes and of lower limb amputation give a much more complex picture of the risk factors compared to an age- and sex-adjustment only. In Table 4, for all Manitobans, the adjusted odds ratio indicates an elevated likelihood of diabetes for the Métis compared to all other Manitobans (aOR = 1.29, 95% CI 1.25-1.34, p < .001; 29% higher), after adjusting for age, sex, geographical location, income, continuity of care, and comorbidities. This is very similar to the relative risk of 1.34 in Table 3, which is only adjusted for age and sex, showing 34% higher risk of diabetes for Métis compared to all others. Despite controlling for other factors, such as living in the North, in lower income areas, and having more physical comorbidities, there is still the persistent effect of being "Métis" and its association with a higher likelihood of having diabetes. This could possibly be genetic susceptibility of the Métis, due to their First Nations ancestry, or to unmeasured factors in our analysis such as dietary considerations [28] or stress levels [29, 30]. Further study would be warranted which includes more comprehensive data sources.

In contrast to the complex modeling of diabetes, the complex modeling of the likelihood of lower limb amputation for those with diabetes shows a dramatically different result when looking at the Métis effect (see Table 5). Table 3 indicates that the age- and sex-adjusted rate of amputation is 49% higher for the Métis compared to all other Manitobans (24.1 vs. 16.2 per 1000, p < .001, RR = 1.49). However, Table 5 shows the more complex regression model with factors such as geography, comorbidity and continuity of care as well as age and sex. This showed no statistically significant difference in the likelihood of lower limb amputation between Métis and all others (aOR = 1.13, 95% CI 0.90-1.40; p = 0.29, NS). Note that the model only has the power to detect an OR of 1.279, so this could be a type 2 error. However, given the dramatic decrease and the small OR after adjustment, even if this were statistically significant it would show that the ethnicity factor of being Métis is clinically not important. So this indicates that factors other than ethnicity have a significant effect on the likelihood of having an amputation once a person has diabetes. Being older or being male, living in an area of lower average household income, or in the Mid and North parts of Manitoba, and having major comorbid physical illnesses all contribute to increased likelihood of amputation. However, having continuity of care shows a beneficial association (aOR = 0.709, p < 0.001). Continuity of care may reflect the difficulties of accessing the same physician for the majority of one's healthcare, especially in rural and northern areas of the province where physician turnover rates may be high. Martens, Bartlett et al. [18] found that provincially, 65.4% of Metis and 69.1% of all other Manitobans received over 50% of their care from the same physician in a two-year period of time of 2005/06 to 2006/07. However, this was much lower in the North area of the province (Métis 58.7%, all other Manitobans 57.2%), and particularly in the RHA of Burntwood (Métis 47.3%, all other Manitobans 47.3%). For diabetes in particular, consistent management and follow-up may prevent adverse outcomes.

So the "Métis" effect of higher amputation rates may be explained by where the Métis live, their lower income levels, their additional burden of comorbidity, or the lower percentage receiving continuity of care. It is important to note, however, that even in the Métis-only analysis, higher average household income level was associated with lower amputation rates, so social policy must be considered. As well, knowing that Métis living in rural and northern remote areas have difficulty accessing the healthcare system, it is important to do further study into healthcare access issues. A previous ecologic study by Martens et al. [26] has shown that diabetes prevalence was highly associated with socioeconomic status amongst First Nations Tribal Council areas of Manitoba, but lower limb amputation was associated with lower rates of access to specialist care (and not with socioeconomic status). This study of the Métis was analyzed at the individual person level, not the aggregate area level, but we see similar findings as to the importance of the healthcare system in being associated with lower rates of amputation for people living with diabetes.