Background Previous studies have observed higher mortality among deprived individuals with type 2 diabetes mellitus (T2DM) than among people with T2DM from more affluent backgrounds. This differential mortality may reflect, inter alia, socially-patterned variation in levels of comorbidity (clinical conditions additional to diabetes). The study reported here sought to establish whether socio-economic status (SES) and comorbidity exert independent effects on mortality at one year after diagnosis with T2DM.
Methods Data for 71,334 men and 57,853 women diagnosed with T2DM in Scotland at ages 25 and above during 2004–2011 were obtained from a national diabetes register. Hospital discharge records covering the period 20 years prior to T2DM diagnosis were used to derive patients’ historic burden of comorbidity as defined by the Charlson Comorbidity Index (CCI), an established measure of comorbidity. Socio-economic status was represented by the Scottish Index of Multiple Deprivation (SIMD), a small-area based measure of SES. A logistic regression model was fitted in which survival at one year after T2DM diagnosis was predicted initially by age and CCI; the model was then augmented with the addition of SES (represented by quintiles of the distribution of SIMD). Values of the discrimination index (representing the area under the Receiver Operating Characteristic [ROC] curve) were obtained for models without and with SIMD quintile, to establish whether addition of SES materially improved the prediction of one-year mortality. Separate models were fitted for each sex.
Results There were 1703 deaths in men and 1502 deaths in women. Proportions of men with no comorbidities ranged from 66.9% (most deprived SIMD quintile) to 73.7% (least deprived SIMD quintile). Corresponding proportions for women were 67.3% (most deprived) and 74.2% (least deprived). For men, the area under the ROC curve (AUC) was 0.804 for the baseline model (no SES), and 0.806 for the model augmented with SES. Corresponding values for women were 0.821 (baseline model) and 0.822 (augmented model). For both sexes, SES exhibited a significant effect on one-year mortality in the augmented model: odds ratios for the contrast between the least deprived and most deprived quintiles of SIMD were 0.701 (95% confidence interval: 0.590, 0.832) for men and 0.766 (0.640, 0.917) for women.
Conclusion Prediction of one-year mortality in people with T2DM was only marginally improved by the addition of SES (relative to the effects of age and of comorbidity). Previously-observed social gradients in mortality among people with diabetes may partly reflect unmeasured effects of comorbidity.