Study selection

Identifying duplicate records

We combined all database searches and then removed all duplicate records. Duplicates were matched using a two step process. First we used our own software pydedup (that can be found here) to automate the majority of the process to remove duplicates. All publication titles were stripped of punctuation and whitespace with only unique titles being returned. Second we imported the remaining records into the reference software Zotero for manual matching of remaining duplicates. This enabled us to identify the remaining small number of close matches e.g. records from one database that contained mathematical symbols that had a representation in words from another database. We manually merged the duplicates in this second stage.

Abstract, title and keyword screening

The titles, abstracts and keywords of all unique records were then screened by one member of the team using the reference management software Zotero. We tagged each record as a ‘Yes’ or ‘No’. Excluded records were re-reviewed in a second pass to mitigate the risk of missing key papers.

Articles were included in the second stage of review and data extraction, if the title, abstract or keywords indicated that health care or patient based DES model was used in an applied setting, cost effectiveness study, or methodology study. We included hybrid DES models; for example those studies that hybridised agent based simulation and DES.