Temporal deep learning framework for retinopathy prediction in patients with type 1 diabetes.

The adoption of electronic health records in hospitals has ensured the availability of large datasets that can be used to predict medical complications. The trajectories of patients in real-world settings are highly variable, making longitudinal data modeling challenging. In recent years, significant progress has been made in the study of deep learning models applied to time series; however, the application of these models to irregular medical time series (IMTS) remains limited. To address this issue, we developed a generic deep-learning-based framework for modeling IMTS that facilitates the comparative studies of sequential neural networks (transformers and long short-term memory) and irregular time representation techniques. A validation study to predict retinopathy complications was conducted on 1207 patients with type 1 diabetes in a French database using their historical glycosylated hemoglobin measurements, without any data aggregation or imputation. The transformer-based model combined with the soft one-hot representation of time gaps achieved the highest score: an area under the receiver operating characteristic curve of 88.65%, specificity of 85.56%, sensitivity of 83.33% and an improvement of 11.7% over the same architecture without time information. This is the first attempt to predict retinopathy complications in patients with type 1 diabetes using deep learning and longitudinal data collected from patient visits. This study highlighted the significance of modeling time gaps between medical records to improve prediction performance and the utility of a generic framework for conducting extensive comparative studies.

Characteristics of newly diagnosed type 1 diabetes in paediatric and adult population from Reims University Hospital, France from 1997 to 2019.

French health insurance data showed that the incidence of type 1 diabetes mellitus (T1DM) in children increased over the years to 2015. The objective of our study was to assess the evolution of the number of incident cases of paediatric and adult type 1 diabetes in our institution, and to describe their clinical presentation and its evolution. All patients with T1DM managed at diagnosis at Reims University Hospital between 1997 and 2019 were included. The clinical and biological data were extracted from the Champagne-Ardenne Diabetes Network database. Included were 847 patients with a median age of 10.3 years. Diagnosis was established in 71% of cases before 15 years, 7.4% after 35 years. The number of newly diagnosed cases was 3.6-times higher in 2019 compared to 1997. Ketoacidosis, the frequency of which decreased with age (P < 0.0001), revealed diabetes in a total of 32% of cases and in 46% of children under 5 years. It was more severe in children than in adults (P = 0.03), and its frequency increased over the study period. Hypotrophy was found in 23% of children under 15 years of age, and was more pronounced before 5 years of age, with no improvement over time. We saw an increase in the frequency of obesity or overweight among adults. Our study showed an increase in incident cases of diabetes in our hospital that continued over time for both children and adults. Clinical features at diagnosis deteriorated during this period for those under 15 years of age with an increase in ketoacidosis frequency.

Early predictors of diabetic retinopathy in type 1 diabetes: The Retinopathy Champagne Ardenne Diabète (ReCAD) study.

AIMS: To determine the relationship between early markers of diabetes control and diabetic retinopathy (DR) in type 1 diabetes. METHODS: A historic cohort study was conducted on 712 patients from the CARéDIAB database. HbA1c and usual metabolic parameters were measured one year after diagnosis of diabetes. First occurrences of severe hypoglycemia and ketoacidosis during follow-up were selected as time-dependent markers of diabetes control. Data were analyzed in a Cox model using SPSS software to predict DR with significance level at p-value <0.05. RESULTS: In multivariate regression, any diabetic retinopathy was predicted by HbA1c (HR = 1.38; CI = 1.25-1.52; p < 0.0001), severe hypoglycemia (HR = 3; CI = 1.99-4.52; p < 0.0001), ketoacidosis (HR = 1.96; CI = 1.17-3.22; p = 0.009), and age at diagnosis (HR = 1.016; CI = 1.002-1.031; p = 0.02). Proliferative DR was predicted by HbA1c (HR = 1.67; CI = 1.51-1.79; p < 0.0001), severe hypoglycemia (HR = 3.67; CI = 2.74-5.25; p < 0.0001), and ketoacidosis (HR = 2.37; CI = 1.56-3.18; p < 0.0001). CONCLUSION: This study shows that the failure to achieve diabetes control after the first year of diagnosis as well as early episodes of acute diabetes complications may contribute to the occurrence of diabetic retinopathy in type 1 diabetes patients.