Work's been pretty busy. Progress since last blog entry? Have I become a machiiine? Have I become angry? Not quite yet but I have been getting more productive. That or there's just so much work that I have no choice to be....
Still getting mindsplit between the two jobs but it's very engaging and I'm never bored, so yay! Next to my original workload, helping with two grants, clinical data collection, document validation, and patient recruitment, I've been learning how to use SQL Server Integration Services, used for data integration and workflow applications and dabbing in MATLAB for data analysis.
From a high level, dealing with data at UCSF and Epic isn't easy. It's kind of like a game of telephone:
- Hospital staff collect and input the data into Epic as they observe them. If the data is not structured, they will enter free text.
- The data is cached in Epic. The software and data model is designed by software engineers.
- Cached data is sent to a relational database (SQL). Which can then be copied into separate databases. The SQL database and data structure is adapted to cached data by software engineers and data scientists.
- Clinicians request data reported in a way they can interpret and analyze.
- Analysts (programmers, computer scientists, data scientists etc) extract data from SQL or other database sources.
- Research assistants analyze the data extractions.
- Clinicians improve or change their practice based on data analysis.
For this to work smoothly, a good understanding of what happens at every part of this pathway is necessary. It is important for everyone to understand data flow and how things get captured clinically and electronically. But providers in a busy large research hospital do not always have the bandwidth to be mindful of data methods. In the moment of service, patients should always come first.
How do we put everybody on the same page?
The easiest way is to change the way hospital staff interfaces with data input and output methods in the first step. Increasing the intuitiveness of graphical user interfaces (GUI) and the data feedback loop will:
- Increase staff awareness on the importance of meaningful data input
- Increase staff appreciation for data output
- Increase the need for standardizing methods of other types of data. (For example, images need to be stored and interpreted.)
What other solutions are there?