Aggregating diagnostics provided by Cepheid's GeneXpert machines
Together with InSTEDD, we developed a platform for centralizing the information provided by the GeneXperts, and providing comprehensive visualizations for the different stakeholders involved: medics analyzing test results trends, engineers involved in maintenance of the machines, and trainers interested in tracking human errors.
The platform allows multiple roles of access to the information, as well as segmentation by the different end-institutions who manage the GeneXperts at the local level. Therefore, all information generated by the devices is centralized in a portal which allows global-level analysis, yet limiting access at different levels due to the sensitiveness of the information being managed.
Our work started by implementing a pilot in South Africa with NHLS and in India with FIND, tracking MTB cases; followed by a Usability Workshop at South Africa with representative users from NHLS and Cepheid to obtain critical feedback on both their needs and their feelings towards the application. The pilot application is now home to over 2.5 million tests, being one of the largest real-time repositories of automated diagnostics private data.
We continued the development of the platform, which has now grown to a complete solution including a Java client, the XpertReporter, which manages tests upload, LIS data recollection and additional data input; as well as a highly optimised communication and processing layer separated from the web frontend. Additional features, such as stronger reporting capabilities, more fine-grained information sharing options, visualizations tailored to each user’s needs, and improved performance for large-scale deployments, were also added.
While designing future versions of the RemoteXpert platform, there is also a new version for tracking Ebola cases being developed for deployment in West Africa.
Overview of system performance: cartridge types, evolution paired against other laboratories and a few success stories.
Interactive dashboard where every table with demographics is, on itself, a filter over the rest.
The scatterplot matrix should help see correlation between different diseases.
Visualization of where laboratories are located through a map.
Request to application for access to test result information from other organizations, done through carefully controlled permissions.
The web front end is developed in Ruby on Rails 3.2, using Google Visualization API for client-side interactive charts; while the processing layer is an Erlang OTP application. The XpertReporter client is Java-based, using a local encrypted SQLite DB. Relational server data is stored in a MySQL master/slave setup, and is indexed by an ElasticSearch cluster for fast queries and aggregations.