Diagknowzit: Data Mining For Social Impact
For hospitals in resource poor settings, implementing an electronic medical record (EMR) system is a necessary — though daunting — undertaking. Since large scale enterprise systems are beyond the reach of most organizations, a thriving open source platform, OpenMRS, emerged in the 2000s as a solution to this informatics need. On the one hand, an EMR system must be thorough and complete, but at the same time, it must be easy for all people to use — especially those at varying levels of medical knowledge and technical skill.
To maximize usability, OpenMRS enables data entry technicians to enter diagnoses into the system as uncoded data, instead of using the formal medical coding system — if they so choose. While this makes the system more accessible, it also creates a bottleneck later on because someone (usually a doctor) must review the diagnosis and code it manually. This work is tedious, and also deprives the doctor of time he or she could be using more effectively, i.e. to save more lives.
Our project is to build a recommendation engine into the data entry module of the OpenMRS system. The engine will be implemented as a web application extension to OpenMRS, and will function similar to Google's "Did you mean...?" feature. We plan leverage our knowledge of machine learning to build the engine, while at the same time being mindful of the limited computational resources that are available in practice. We are excited to leverage the power of data mining algorithms on behalf of non-profits, a space where such methods remain under-utilized.