Development and validation of the DIabetes Severity SCOre (DISSCO): a retrospective cohort study
The prevalence of type 2 diabetes (T2DM) is rapidly increasing worldwide and the importance of assessing diabetes severity is well recognised. However, validated type 2 diabetes severity measures derived from medical data are lacking with limited applications in clinical practice. No previous studies have compared the predictive value of severity measures to HbA1c measurements. This study aimed to: develop type 2 DIabetes Severity SCOre (DISSCO) for predicting hospitalisation and mortality, assess the added clinical utility of DISSCO for predicting outcomes beyond that achieved using models incorporating demographic and clinical variables including HbA1c, and validate the developed severity score using an independent validation dataset.
A retrospective cohort study using routinely-collected electronic health records held by the Clinical Practice Research Data link (CPRD) between 2007 and 2017. People aged ≥35years, diagnosed with T2DM, and registered in English general practices were included. The study population was randomly divided into 80% and 20% of the cohort as the training and validation datasets, respectively. Baseline and longitudinal severity scores, were developed based on diabetes-related complications and cardiovascular disease. Measured outcomes were all-cause mortality, first hospital admission episode for any cause, due to diabetes including hypoglycaemia, or cardiovascular disease and cardiovascular procedures.
A cohort of 139,626 people with T2DM (111,748 in the training dataset and 27,878 in the validation dataset) from 400 general practices was identified. Overall, 32,204 (23%) patients died and 99,951 (72%) patients had at least one hospitalisation during follow-up. In the training dataset, a 1-unit increase in baseline severity score was associated with significantly increased risks for mortality (unadjusted hazard ratio (HR) 1.30, 95% confidence interval: 1.30; 1.31), any hospitalisation (1.17, 1.17; 1.18), and cardiovascular-related hospitalisation (1.59 (1.58; 1.59)). Fully-adjusted models (adjusted for age, gender, HbA1c, ethnicity, and deprivation) reduced the association to 1.13 (1.12; 1.14, AUROC=0.77); 1.09 (1.08; 1.09, AUROC=0.62); 1.45 (1.44; 1.46, AUROC=0.76) for the three outcomes, respectively. The addition of DISSCO to clinical and demographic variables improved the predictive value of models for mortality and hospitalisation outcomes. Compared to the severity score, baseline HbA1c was a weaker predictor for all outcomes, based on AUROC score, except for the risk for hypoglycaemia hospitalisation. Findings were consistent in the validation dataset.Conclusions: higher T2DM severity scores are associated with increased risks for hospital admissions and mortality. The new severity score had higher predictive value than the proxy used in clinical practice, HbA1c.
This new algorithm will be informative to practitioners, focussing where more care is needed, and could stratify clinical management of T2DM patients and support commissioning and public health programmes for people with diabetes. In addition, these findings of our disease-specific severity measure can have important implications for funding future research for grading the severity of other illnesses using routinely-collected data.