Comorbidity or disease co-occurrence is the simultaneous presence of one or more diseases. Application of network analysis to understand disease comorbidity is an active research area.
We build implicit networks using disease comorbidity information available in electronic health records (EHR) datasets. The structural properties from such networks can be combined with individual patient history and population level data to address big data problems.
Using implicit disease co-occurrence networks, we wish to fulfill two objectives:
- Exploring the network to identify interesting patterns
- Incorporating the structural properties of the network into predictive models to improve their prediction performance
Our approach can be applied to analyze a range of relevant problems such as predicting patient re-admissions, treatment costs, length of patient stay and disease predictions.