gms | German Medical Science

Objectives: Periprosthetic joint infections (PJI) represent a primary cause of THA and TKA revisions. PJI, however, is a highly complex disease and every patient is unique. Currently, there is no systematic nor centralised approach to record this high-dimensional patient data for further applications or evaluations. Unformatted storage of information in hospital information systems or Excel files is still the most prevalent way in German hospitals. Nevertheless, such an antiquated assessment does not fully exploit the potential of our systems medicine approach and impedes Artificial Intelligence applications due to lack of data quality and quantity. Thus, innovative, digital analysis tools are urgently needed to further explore causal relations in complex data.

Methods: We developed a database as backend with an easy to use web interface as frontend for data acquisition. The database is designed for usage of experts as well as non-experts and embedded into our high security clinical networks. Therefore, it is accessible for everyone within the network with the respective user rights.

So far, we integrated single-center data for all PJI cases treated at our institution since 2013 (n = 680). The data is stored in a secure set-up and can easily be exported if needed for further analysis. Furthermore, functionality such as mandatory radio buttons and drop-down menus instead of free text boxes assure a structured and complete data acquisition.

Results and conclusion: This study presents the design and architecture of a testable, scalable, and effective web-based application for intuitive data input and analysis, which can for example be used during PJI board meetings for direct digital, structured and fast data acquisition. Moreover, the whole implementation is designed to store data accurately and to capture the state of a patient across time, ensuring privacy and security as well. The interface leads to significant improvement in time for documentation and is less error-prone than common and analogue documentation approaches. For future use, data from partner institutions can also be entered into our database, and Big Data and suitable datasets for Machine Learning applications are a significant step closer.

Figure 1 [Fig. 1]