COSMO-T1D

 

Model Website

Information last updated: July 2022

Participated in following Mt Hood Diabetes Challenge Meetings: 2019 Seoul, Korea. 

Publicly accessible?: The equations of the model have been published in sufficient detail to enable the model to be replicated by other researchers. A software implementation of the model with a user interface is being developed by the University of Oxford in collaboration with the University of Melbourne.

Is the model continuing to be developed?: Yes.

Brief Description:

The COSMO-T1D is a patient-level, probabilistic discrete-time simulation model based on an integrated system of 30 equations for predicting occurrence of diabetes-related complications and progression of risk factors for the complications. Data from the Swedish National Diabetes Register (NDR) were used for model development. The COSMO-T1D consists of 14 parametric proportional hazards models, of which 10 are used to predict probabilities of first and second acute complications (coronary vascular event, stroke, amputation, severe hypoglycaemia and severe hyperglycaemia), three to predict the probability of diagnosis of
three chronic conditions (heart failure, peripheral vascular disease and end-stage renal disease), and one to predict the probability of all-cause mortality. Monte-Carlo methods are used to predict occurrence of the events within a year based on the estimated annual
probabilities. When a coronary vascular event occurs, a multinomial logit model is used to predict if the event is myocardial infarction, percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG). Predictors of the risk equations include time-independent factors (e.g., age at disease onset, sex), time-varying risk factors (e.g., HbA1c, HDL cholesterol, eGFR) and time-varying history of complications (e.g., indicator for occurrence of stroke).

 

The COSMO-T1D uses seven linear regression models to predict progression of continuous risk factors (Hb1Ac, eGFR, BMI, HDL cholesterol, LDL cholesterol, triglycerides and systolic BP) and eight logistic regression models to predict changes in binary risk factors (smoking initiation, smoking cessation, development of microalbuminuria from non-albuminuria, macroalbuminuria becoming microalbuminuria, remission of microalbuminuria, development of macroalbuminuria from non-albuminuria, microalbuminuria becoming macroalbuminuria and remission of macroalbuminuria). Inputs of the model are individuals with pre-defined baseline characteristics (e.g., current age, age at disease onset, HbA1c, history of severe hypoglycaemia, time since last severe hypoglycaemia). Outputs of the model is a longitudinal dataset containing annual values of risk factors and indicators for occurrence of complications and death for each individual. COSMO-T1D allows minimization of first-order uncertainty by implementing Monte Carlo simulation with a large number of replications for each individual. Evaluation of interventions for type 1 diabetes can be conducted by adapting the equations for risk factor progression and/or event risks to simulate its impact on changes in risk factor values and incidence of the complications. For a cost-utility analysis, decrements of health utility and costs associated with the complications can be applied to the model output to estimate quality-adjusted life years and total healthcare costs.

Funding source for model development: 

The study was supported by the National Health and Medical Research Council (NHMRC; grant number 1028335), the Australian Research Council’s Discovery Early Career Researcher Awards scheme (DECRA; grant number DE150100309), and the Australian Research Council’s Centre of Excellence in Population Ageing Research (CEPAR; grant number CE170100005). The Swedish Association of Local Authorities and Regions funds the Swedish National Diabetes Register (NDR) which was used as a data source for model development.

Key Publications:

Tran-Duy A, Knight J, Palmer A, et al. (2020) A patient-level model to estimate lifetime health outcomes of patients with type 1 diabetes. Diabetes Care 43: 1741-1749


Tran-Duy A, Knight J, Clarke P, Svensson AM, Eliasson B, Palmer A (2021) Development of a life expectancy table for individuals with type 1 diabetes. Diabetologia 64: 2228-2236. 10.1007/s00125-021-05503-6