Table 3.

Assessment of a CHEM implementation in youth mental health using transparent, reusable and updatable (TRU) criteria

Criterion	Met?	Detail
T1 Software files are open access	Yes	All source code and testing procedures are available in public GitHub repositories, with each code release persistently available in a Zenodo repository. As the study dataset contains confidential patient records, it was not published. Instead, a synthetic representation of the study dataset is persistently available in a repository in the Harvard Dataverse. Data files to support out of sample application of models are published at the same location
T2 Developer contributions and judgements on appropriate use are easily identified	Partial	All module libraries, programs and datasets are distributed with citation information. Public GitHub repositories detail author code contributions over project development history. Model catalogues persistently available on the Harvard Dataverse describe the predictive performance of models under multiple usage regimes. Each code library is documented with worked examples of how to apply CHEM modules. However, documentation of individual functions and methods are algorithm generated and need more customisation, including more examples of use. Analysis and reporting programs are self-documenting. Sub-routines for generating reports are documented with README files
R1 Programming practices promote selective reuse	Yes	CHEMs are written using both functional and object-oriented paradigms. Code library documentation websites include hypothetical examples of generalisability (applying study algorithm to estimate mapping models from new data with the same predictor and outcome variables) and transferability (adapting study algorithm to develop mapping models from datasets with different predictor variables and outcomes measured with a different utility instrument)
R2 Licenses permit derivative works	Yes	All code is distributed using GPL-3 licenses. Datasets use an amended version of template terms provided by Harvard Dataverse, allowing reuse of data subject to some ethical restrictions (e.g. use in efforts to reidentify study participants is prohibited)
U1 Maintenance infrastructure is in place	Yes	All code is version controlled using git and GitHub, with semantic versioning. Each code library has a specified maintainer, and guidance for potential code contributors is available on the project website
U2 Releases are systematically re-tested and deprecated	Partial	Continuous integration is used for all code libraries, primarily for acceptance testing. Only limited use is made of unit testing. Retired library code is deprecated using tools from the lifecycle R library. Library documentation articles and datasets are also deprecated