Researchers use data mining and network analysis to understand Type 2 diabetes
The University of Sydney announced that its researchers developed a comorbidity network for Type 2 diabetes using advanced data mining and network analysis techniques. The new study was published in the International Journal of Medical Informatics.
Researchers from the University of Sydney believe untapped medical records could be used to predict when a person is at risk of developing type 2 diabetes.
According to lead author of the study Mr Arif Khan, postdoctoral researcher from the Centre for Complex Systems in the Faculty of Engineering and Information Technologies, it is well known that chronic diseases such as type 2 diabetes do not occur in isolation, and have a shared set of causes common to many other diseases and disorders.
“Chronic diseases like type 2 diabetes progress slowly and, in many cases, patients are unaware of their condition. When they are admitted in to hospital for any incidence, type 2 diabetes often comes up as secondary diagnosis. This makes the overall treatment plan more complex, increasing ‘Length of Stay’ in hospital and cost,” Mr Khan added.
As such, the aim of the research is to understand how health trajectory differs between type 2 diabetes and non-type 2 diabetes patients with the help of comorbidity network.
Together with Capital Markets Cooperative Research Centre (CMCRC), the researchers analysed 1.4 million admission records from nearly 1 million de-identified patients using routinely collected administrative healthcare data. The dataset used for this study was provided by data custodians Capital Markets CRC Ltd. and HAMBS Ltd.
The dataset was then filtered and sampled to obtain a cohort of 2300 diabetics and 2300 non-diabetic patients.
Using advanced data mining and network analysis techniques, the researchers evaluated the health of the cohort and developed a comorbidity network – a way of visualising the health journey of the type 2 diabetes patients.
The network models the relative prevalence of type 2 diabetes comorbidities, meaning additional diseases or disorders co-occurring with a primary disease and their transition patterns, thereby representing the progression of the disease.
In the network image, the size of the nodes (circles) and labels are proportional to the prevalence of corresponding health conditions. Nodes marked in the same colour belong to same clusters of related health conditions. Arcs, in clockwise direction, indicate transition from one health condition to another.
The researchers found that, over time, prevalence of comorbidities in the group of diabetic patients was almost double that of the non-diabetic patients over time, indicating entirely different ways of disease progression.
Study co-author Dr Shahadat Uddin from the Centre for Complex Systems and the Charles Perkins Centre said the comorbidity network could help healthcare providers proactively identify patients at higher risk of developing chronic disease.
“By using existing administrative healthcare data – which are routinely collected but often neglected in health research – we have been able to understand the ‘disease footprints’ left by chronic patients,” Dr Uddin said.
“These insights can subsequently help healthcare providers to better understand high-risk diseases and to formulate appropriate preventive health policies,” added study co-author Dr Uma Srinivasan from the CMCRC.
The researchers predicted their comorbidity network could also be used effectively to monitor other chronic diseases, such as cardiovascular diseases.