Methods of analysing patterns of multimorbidity using network analysis: a scoping review

Aim: The purpose of this scoping review is to summarise the techniques used for network analysis of multimorbidity to inform development of a standard methodology.

Subject and method: Multimorbidity is the presence of two or more health conditions. Many techniques have been applied to investigate the relationships between health conditions in people with multimorbidity, including network analysis. Network analysis finds patterns in a complex set of variables and displays them in a visualization or network map. There is currently no recommended standard method for calculating and displaying networks in multimorbidity. Studies were included if they used a network analysis method to investigate the relationships between multiple illnesses in a population sample, without reference to a specific health condition.

Results: A total of 15 articles were found meeting inclusion criteria of which six were published in 2020. Not all articles defined multimorbidity or referred to multimorbidity. Several techniques were used to measure the association between conditions, including correlation coefficients and observed/expected ratios. R and Stata were the most popular software packages used. Some articles did not define their measure of association or mention which software package they used.

Conclusion: There is a need for a greater level of detail and transparency when reporting methodology and software used for network analysis of multimorbidity, and a formally recognised definition of multimorbidity. Further investigation into the most appropriate measure of association for use in network analysis in the context of multimorbidity is required to improve techniques investigating interactions between multiple health conditions.