Data were obtained from 100 of the 101 general practices. One practice declined access. Diabetes age-standardized prevalence (European standard population) in 2004 was 5.9/100 and in 2009 7.3/100. Between 2004 and 2009 there were 37,297 people with a Read code for type 2 diabetes. Diabetes treatment was recorded in one or more years in 31,697, (85%) and the final sample consisted of 24,111, (76%) who satisfied the age inclusion criteria. Self-reported ethnicity was recorded in 23,483, (97%) of whom 5,206, (22%) were White, 13,633, (58%) South Asian, 3,923, (17%) Black African/Caribbean and 721, (3%) other ethnic groups.
Townsend score was recorded in 23,995, (>99% patients), higher scores indicating more deprivation. Only 2% of patients fell in the two least deprived Townsend groups and the three least deprived groups were combined. Patients were categorized as; ‘Least deprivation’ if their Townsend score was less than 3.93 (N = 4,900, 20%), ‘High deprivation’ if their score was above 6.75 (N = 7,341, 31%) and ‘Medium deprivation’ (11,754, 49%) otherwise. There was a higher proportion of South Asians in the ‘High deprivation’ group. Blood pressure was recorded in 98% of the cohort, BMI 95% and cholesterol 95%. The proportion of patients with HbA1c recorded increased every year; 79% of patients had an HbA1c value recorded in 2004 and 87% of patients had an HbA1c record in 2009.
Table describes the characteristics of people with and without HbA1c records for the entire five year cohort in whom recording by ethnic and social group was similar. Table shows the demographic variables for the entire five year cohort. Of the total sample over the 5-year period 27% were on ‘Metformin Only’, 50% on ‘Combined oral’ and 23% on ‘Insulin’. In the least deprivation group 51% had HbA1c controlled, compared to 45% in the high deprivation group.
Proportion of each category with an HbA1c value recorded 2004–9
Demographic and clinical characteristics by ethnicity and Townsend category 2004–9
From 2004 to 2009 there was little change in the proportion of patients on each treatment; 26% of patients were on ‘Metformin Only’ in 2004, compared to 29% in 2009. 52% of patients were on ‘Combined oral’ in 2004 compared to 48% in 2009. 23% of patients were on Insulin in both 2004 and 2009. White people (69%), were less likely to be on intensive diabetes treatment (either ‘Combined oral’ or ‘Insulin’) than South Asian (75%) and Black African/Caribbean people (73%) but South Asian people were less likely to be on insulin (21%), than White people (27%). A chi-squared test indicated these treatment differences were significant (P < 0.001).
Between 2004 and 2009, unadjusted mean HbA1c for White people declined by 0.4% from 8.2% to 7.8%; for South Asian and Black African/Caribbean people it declined by 0.5% from 8.5% to 8.0%. Similarly, the proportion of people with HbA1c controlled to 7.5% or less, increased by 12% in White people (from 44% in 2004 to 56% in 2009), by 14% in South Asian people (from 34% to 48%) and by 15% in Black African/Caribbean people (from 38% to 53%).
Tables and Appendix 1 (available at http://jrsm.rsmjournals.com/lookup/suppl/doi:10.1258/jrsm.2012.110289/-/DC1
) show the parameter estimates from the multilevel models used to derive adjusted HbA1c mean and HbA1c control outcomes. PCT was dropped from the model as it was not significant (P
= 0.095). There was significant evidence to justify inclusion of age, gender, year, Townsend score, serum cholesterol, smoking status, ethnic group and treatment in both models. BMI did not significantly influence HbA1c (P
= 0.227) but was retained in both models as a variable of interest. The ethnicity*year interaction was significant in both models indicating that the difference between ethnicities in mean HbA1c and HbA1c control changed between 2004 and 2009. There was no interaction between ethnicity and social deprivation (P
= 0.475). Ethnic group was more strongly associated with HbA1c than social deprivation with regression estimates of 0.36 and 0.32 for South Asian and Black ethnic groups compared to 0.02 for Townsend score.
Predictors of HbA1c using a linear multilevel regression model
In this adjusted analysis, men, smokers and those with high deprivation had higher HbA1c and lower odds of controlled HbA1c. Increased treatment intensity ‘Combined oral’ or ‘Insulin’ treatment were associated with a higher mean HbA1c, 0.4% and 1.1% respectively, and lower probability of controlled HbA1c relative to ‘Metformin only’ treatment.
In 2004, mean HbA1c was significantly lower in White people relative to South Asian people; difference = 0.36%, P < 0.001 and Black African/Caribbean people; difference = 0.32%, P < 0.001. Similarly in 2004, HbA1c controlled to 7.5% or less was more likely in White people than South Asian people (OR 2.1, 95% CI: 1.8, 2.5) or Black African/Caribbean people (OR 1.7, 95% CI: 1.4, 2.1).
The ethnicity*year interaction (Table ) shows that from 2004 to 2009 the difference in mean HbA1c between White and Black African/Caribbean people decreased significantly by 0.2% (P < 0.001) from 0.32% to 0.12%. In addition, the odds of a White person having controlled HbA1c relative to a Black African/Caribbean person was significantly reduced to 1.3. The difference in mean HbA1c between White and South Asian people also decreased significantly by 0.11% (P = 0.007) from 0.36% to 0.25%. However, the odds of a White patient having controlled HbA1c in 2009 relative to a South Asian patient was not significantly different to the odds in 2004 (P = 0.118).
Illustrative graphs in Figure show the expected mean HbA1c and probability of having HbA1c controlled each year by ethnic group (derived for a non-smoking male on ‘Metformin only’ using a mean age of 53 years and serum cholesterol 4mmol/l). The graphs show a reduction in HbA1c for all ethnic groups since 2004, most marked in earlier years, and a small reduction in the difference in mean HbA1c and HbA1c control between White and other ethnic groups.
Estimated mean HbA1c (a) and probability of controlled HbA1c(b) by ethnic group 2004–9
Social deprivation remained independently associated with HbA1c after adjustment for other factors including ethnic group. As an illustrative example, Table shows the expected mean HbA1c and probability of having controlled HbA1c for a white non-smoking male patient in 2009 on ‘Metformin only’ in the least, medium and high social deprivation groups. The table shows that high social deprivation increases mean HbA1c by 0.1% (1 mmol/mol) and reduces the probability of having controlled HbA1c by 3% relative to being in the least deprivation group. Similarly Table shows the difference in expected mean HbA1c and probability of having HbA1c control between each ethnic group. Table shows the relationship between HbA1c, ethnic and social group. The non-significance of the ethnicity*Townsend interaction means an increase in deprivation effects similar changes in the HbA1c levels of each ethnic group.
Predicted mean and controlled HbA1c in 2009 for each Townsend category
Predicted mean HbA1c and controlled HbA1c in 2009 for each ethnic group
Predicted mean HbA1c for each ethnic and social group in 2009
There was concern that the sample age range selectively excluded more South Asian people with diabetes aged 18-34 years. Therefore the HbA1c model was rerun including all 2612 patients of these ages of whom 11% were White, 76% South Asian, 13% Black African/Caribbean and 17% were least deprivation, 47% medium and 36% high deprivation. The statistical models indicated ethnic group (P < 0.001) and social group (P < 0.001) were still significant predictors of HbA1c and the difference in mean HbA1c between ethnic groups still decreased significantly (P < 0.001) over the study period.