Socio-Economic Status and Hemoglobin A1c Levels
Socio-Economic Status and Hemoglobin A1c Levels
We studied the association of Hgb A1c levels with annual income, material and social deprivation in a Canadian primary care population. We found an inverse relationship between SES and Hgb A1c and very small differences in mean Hgb A1c levels between the most materially deprived populations and those with lesser deprivation. A threshold for clinically significant differences in Hgb A1c in the Canadian setting has been agreed upon as being 10%. In our study, this would have meant differences in Hgb A1c of 0.6% or more; we found differences that were an order of magnitude smaller. These differences were no longer statistically significant in three out of four SES models after adjustment for other factors known to be associated with an increase in the incidence of diabetes.
Previous studies have found an association between lower income, lower education level, lower employment grade and increasing Hgb A1c in persons without diabetes. A strength of this study is the fact that we were able to study a larger primary care sample, and adjusted for multiple factors associated with a greater risk of incident diabetes. As well, we used validated indices aggregating several aspects of deprivation.
The literature was unclear as to which elements of deprivation might be associated with Hgb A1c levels; we therefore used several validated indices of deprivation (income, material, social, combined). This is similar to the approach taken in a recent study to study obesity and deprivation. After adjustment, only material deprivation was statistically associated with Hgb A1c levels, although the difference was small and not clinically significant.
Neighborhood deprivation may be associated with increases in cardiometabolic risk factors and levels of obesity. Several of these risk factors are associated with an increased risk of incident diabetes. These could explain the higher rates of incident diabetes observed in deprived areas, rather than neighborhood poverty independently affecting Hgb A1c.
We found associations between increasing Hgb A1c and factors included in diabetes risk calculators, such as increasing BMI, increasing blood pressure, increasing fasting blood glucose, age, gender, family history and dyslipidemia. This supports the use of these factors to assist in diabetes screening decisions.
We also noted an increase in the uptake of Hgb A1c screening in patients age 45 or more, from 20% in our previous study (immediately prior to the release of guidelines recommending this test in persons at risk of developing diabetes) to 32% in the current study, two years after the release of the new guidelines.
In the Canadian context, citizens have universal coverage for health care. The finding that lower SES was not an independent risk factor for elevated Hgb A1c in the patients of family physicians in this study may be generalizable to clinical settings where SES is not a barrier to accessing primary care. However, this lack of association should not be interpreted to mean that interventions to reduce the risk of diabetes in lower SES neighbourhoods are inapplicable at the population level.
We used an ecological approach to measure deprivation as we could not determine deprivation directly at the individual level. However, the measures we used have been validated and have been extensively employed in other studies.
We had several data limitations. Waist circumference is an important predictor of diabetes. We could not include this factor as 75% of our sample population did not have a waist circumference recorded in the EMR during the three years of interest. Some of the records lacked precision as to which relative had a history of diabetes; in other words, we did not know whether the family history was in a first degree relative. For this study, family history of diabetes was defined as having a recorded history of diabetes in any blood relative. Ethnicity is an important predictor of diabetes, but is poorly captured in the EMRs we used and could not be included. Similarly, EMR data on physical activity levels, smoking history, and diet are incomplete, and could not be used in this study. Postal code was missing for a small percentage of patients(less than 2%); while some of these patients may be homeless, the proportion is not large enough to invalidate our study results. Only 37.7% of the sample was male. Male patients represent 36% of the CPCSSN patient population in Toronto; the percentage of males in study population was similar to that of the source primary care population.
This was an observational study using EMR data, and there were likely systematic differences between patients tested and not tested using Hgb A1c that could impact the generalizability, but not internal validity, of these findings. In our previous study, patients with factors associated with a higher risk of incident diabetes were more likely to have the test done. As well, persons living in the lowest income quintile neighborhoods had a higher adjusted odds ratio of having a screening Hgb A1c test done than those living in the highest income quintile (OR 0.63). However, selectively testing patients with more risk factors, which tend to cluster in poorer neighborhoods, may lead to bias towards falsely positive differences instead of our generally negative results. Lastly, our study reflects conditions for persons living in a largely urban setting in southern Ontario; factors affecting neighborhood deprivation may differ in other settings.
Discussion
We studied the association of Hgb A1c levels with annual income, material and social deprivation in a Canadian primary care population. We found an inverse relationship between SES and Hgb A1c and very small differences in mean Hgb A1c levels between the most materially deprived populations and those with lesser deprivation. A threshold for clinically significant differences in Hgb A1c in the Canadian setting has been agreed upon as being 10%. In our study, this would have meant differences in Hgb A1c of 0.6% or more; we found differences that were an order of magnitude smaller. These differences were no longer statistically significant in three out of four SES models after adjustment for other factors known to be associated with an increase in the incidence of diabetes.
Previous studies have found an association between lower income, lower education level, lower employment grade and increasing Hgb A1c in persons without diabetes. A strength of this study is the fact that we were able to study a larger primary care sample, and adjusted for multiple factors associated with a greater risk of incident diabetes. As well, we used validated indices aggregating several aspects of deprivation.
The literature was unclear as to which elements of deprivation might be associated with Hgb A1c levels; we therefore used several validated indices of deprivation (income, material, social, combined). This is similar to the approach taken in a recent study to study obesity and deprivation. After adjustment, only material deprivation was statistically associated with Hgb A1c levels, although the difference was small and not clinically significant.
Neighborhood deprivation may be associated with increases in cardiometabolic risk factors and levels of obesity. Several of these risk factors are associated with an increased risk of incident diabetes. These could explain the higher rates of incident diabetes observed in deprived areas, rather than neighborhood poverty independently affecting Hgb A1c.
We found associations between increasing Hgb A1c and factors included in diabetes risk calculators, such as increasing BMI, increasing blood pressure, increasing fasting blood glucose, age, gender, family history and dyslipidemia. This supports the use of these factors to assist in diabetes screening decisions.
We also noted an increase in the uptake of Hgb A1c screening in patients age 45 or more, from 20% in our previous study (immediately prior to the release of guidelines recommending this test in persons at risk of developing diabetes) to 32% in the current study, two years after the release of the new guidelines.
In the Canadian context, citizens have universal coverage for health care. The finding that lower SES was not an independent risk factor for elevated Hgb A1c in the patients of family physicians in this study may be generalizable to clinical settings where SES is not a barrier to accessing primary care. However, this lack of association should not be interpreted to mean that interventions to reduce the risk of diabetes in lower SES neighbourhoods are inapplicable at the population level.
Limitations
We used an ecological approach to measure deprivation as we could not determine deprivation directly at the individual level. However, the measures we used have been validated and have been extensively employed in other studies.
We had several data limitations. Waist circumference is an important predictor of diabetes. We could not include this factor as 75% of our sample population did not have a waist circumference recorded in the EMR during the three years of interest. Some of the records lacked precision as to which relative had a history of diabetes; in other words, we did not know whether the family history was in a first degree relative. For this study, family history of diabetes was defined as having a recorded history of diabetes in any blood relative. Ethnicity is an important predictor of diabetes, but is poorly captured in the EMRs we used and could not be included. Similarly, EMR data on physical activity levels, smoking history, and diet are incomplete, and could not be used in this study. Postal code was missing for a small percentage of patients(less than 2%); while some of these patients may be homeless, the proportion is not large enough to invalidate our study results. Only 37.7% of the sample was male. Male patients represent 36% of the CPCSSN patient population in Toronto; the percentage of males in study population was similar to that of the source primary care population.
This was an observational study using EMR data, and there were likely systematic differences between patients tested and not tested using Hgb A1c that could impact the generalizability, but not internal validity, of these findings. In our previous study, patients with factors associated with a higher risk of incident diabetes were more likely to have the test done. As well, persons living in the lowest income quintile neighborhoods had a higher adjusted odds ratio of having a screening Hgb A1c test done than those living in the highest income quintile (OR 0.63). However, selectively testing patients with more risk factors, which tend to cluster in poorer neighborhoods, may lead to bias towards falsely positive differences instead of our generally negative results. Lastly, our study reflects conditions for persons living in a largely urban setting in southern Ontario; factors affecting neighborhood deprivation may differ in other settings.