Performing surveillance for healthcare-associated infections (HAI) is not easy. Infection Preventionists do their best at applying sometimes cumbersome definitions. In most hospitals, the surveillance process is at least partially done using electronic medical records and a third-party abstraction software. However, the work is not over when identifying HAIs. There are downstream steps including packaging the data and uploading them to the National Healthcare Safety Network (NHSN). Hospitals may prefer to maintain some of their data locally and report HAI rates for internal consumption using the local source. This approach creates two sources of data: a) data stored locally and b) data stored in NHSN. While, ideally, they mirror each other; differences between these two sources are common. Once data are in NHSN, hospitals download summary statistics including standardized infection ratios (SIRs) for benchmarking purposes. Lastly, those downloads are entered and processed in a visualization software (MS Excel, Tableau, Power BI, etc.) where data are shaped into useful information and presented to key stakeholders including frontline staff.
Given the number of data transfer steps and manipulations/reformatting of the data, it is not surprising that accuracy can be compromised. Below, we describe how we decreased the number of steps in the data flow to improve accuracy.
We started by integrating our electronic medical record and abstraction tool (currently, we perform abstraction within the electronic medical record). We also made NHSN our main data repository, thus, avoiding having two sources of data. We download raw files from NSHN rather than summary statistics: the raw data is then manipulated using programming language (code) that remains stable from month to month, eliminating manual manipulation. Downloading raw files also allows us to calculate metrics not readily available within NHSN, see Surgical Infections pt 2 – Comparing Surgeons.
Validation files are created for review by Infection Preventionists to ensure data was not distorted during the trip from EMR to NHSN and back. Once data are validated, information is disseminated to hospital stakeholders.
While the proposed data flow model can be resource intensive, in the long run, it has decreased the time spent doing surveillance (by integrating the EMR and abstraction tool), stopped manual manipulation (by downloading raw files and using code), and overall nearly eliminated errors in the information reported to both NHSN and frontline staff.