Using Rmarkdown to Generate Standardized Quality Improvement Reports

Monday, June 20, 2016: 11:30 AM
Tubughnenq' 5, Dena'ina Convention Center
Eric Bakota , Houston Department of Health and Human Services, Houston, TX
Avi Raju , Houston Department of Health and Human Services, Houston, TX
BACKGROUND: Health departments have datasets that can be used for performance management and quality improvement (QI). Management reports at the City of Houston Health Department (HHD) are created on an ad hoc basis and when regularly produced require significant investments in manual data abstraction, cleaning, and interpretation. Each step has the potential to introduce errors in the final product. HHD explored using RMarkdown as a tool to improve a QI report that gauged timeliness and completeness of surveillance investigations. The former procedure used traditional Microsoft Office software to generate the report on a monthly basis. This traditional report was time consuming and prone to user errata. RMarkdown is a tool that will allow for full automation of reports on a schedule or as-needed basis.  

METHODS: HHD has migrated from manual performance reports of epidemiologists and surveillance investigators towards an automated reporting system. These reports include a variety of performance and quality metrics, including timeliness and completion of surveillance investigations. The manual report was created with Office products (Access, Excel, and Word). The automated system uses RMarkdown, an open source tool that leverages the statistical and graphical power of R with the documentation style of HTML and/or LaTeX. The original, manually produced QI report was compared with the automated report for three months.

RESULTS: The 6-8 page QI report manually created with Microsoft Office required approximately 6 hours per month of an informatician’s time. To alleviate the reporting burden, the same QI report was created in RMarkdown and compared to the MS Office version. The RMarkdown version took approximately 30 hours to create. After the initial investment, however, each month’s report requires minimal time (less than 1 hour for data cleaning) and produced no user errors. The manually generated report, however, regularly contained user errata, both typographical and statistical. Time savings are realized after 6 iterations of reports.

CONCLUSIONS: RMarkdown and automated reporting in general can uncover new efficiencies in the workplace, allowing epidemiologists and informaticians to do more with less. These automated reports will have several benefits: a professional layout, fewer errors, more utility of existing data, reduced labor needs, and consistent documentation for PHAB Accreditation. Additionally, health departments that become more familiar with R as the primary statistical tool will be able to migrate away from other high-cost tools and reallocate resources.