Skip to main content

Diabetes Mellitus

Description

Illnesses like infections, cold, influenza and so on in type 1 diabetes mellitus (T1DM) patients, can compromise the daily patient administered diabetes treatment. This in turn may result in fluctuating blood glucose concentrations, especially hyperglycemia for prolonged periods, which over time can cause serious late complications. The illness prediction project at Tromsø Telemedicine Laboratory aims to construct a prediction model that, through use of patient observable parameters, for example, blood glucose, insulin injections and body temperature, can significantly identify risk of developing illnesses, before onset of symptoms and before illness onset.

Such a model could potentially enable T1DM patients to fight the illnesses, and prepare for an adequate change in the T1DM-management earlier on.

 

Objective

To develop an illness prediction model that can predict illness in T1DM patients before onset of symptoms, using the patient’s observable parameters.

Submitted by hparton on
Description

There is a clear need for improved surveillance of chronic diseases to guide public health practice and policy. Chronic disease surveillance has tended to use administrative data, due to the need to link encounters for an individual over time and to have complete capture of all encounters. Case-detection algorithms generally combine variables found in the data using Boolean operators (i.e., AND, OR, NOT). For example, a commonly used algorithm for DM surveillance requires a patient to have one hospitalization or two physician visits within two years with a diagnostic code for DM. While this approach to defining case-detection algorithms is straightforward, it has limitations. For example, if more than simple combinations of one or two variables are used, then it becomes unwieldy to represent the algorithm and it can be difficult to identity how different variables in the definition contribute to detection accuracy. A multivariable probabilistic case-detection algorithm can address these problems and facilitate exploration of how the multiple variables available from different data sources might improve case-detection accuracy1. In this research, we develop an approach for probabilistic multivariable case-detection and apply the method to a cohort of older adults with known DM status to demonstrate and evaluate the method.

Objective

To develop and validate a multivariable probabilistic algorithm for detecting cases of diabetes mellitus (DM) using clinical and demographic data.

Submitted by knowledge_repo… on