Abstract:
Comorbidity is a co-existence of two or more diseases in a patient. It is an interesting topic in public health. Hence, we are interested to study network analysis of comorbidity in Thai elderly. The focus of the study are to compare and to find appropriate layouts for the network analysis, using the patient aged 60 years and over admission data in 2019 from the National Health Security Office. Our main finding is that there are advantages and disadvantages to defining different patterns and criteria for categorizing interaction patterns for network display. For cases where the study wish to see an overview of the entire network to see which diseases are most common, the prevalence threshold should be set at the low percentile in order for the network to show a large number of nodes. However, this may not be suitable for viewing types of interaction patterns because there may be a lot of edge. For those who want to know which diseases are highly related, should set a high lift threshold because the lift value is indicating the degree of relationship between the pair of nodes. And the appropriate set of parameters of the network is the key to help understand the relationship within comorbidity network. For general purpose, we recommend using the Fruchterman-Reingold layout because the algorithm tends to minimize edge crossings and distributes the vertices evenly in the frame making the nice-looking and easy-to-understand network. When needed, networks from different layouts can be considered together to provide further insights.