Multicentric quality benchmarks
Healthcare organizations are constantly striving to deliver high-quality care. Collecting and sharing healthcare data is crucial to improve patient outcomes, advancing medical research, and optimizing healthcare systems which can result in greater efficiency and cost savings for healthcare organizations. It also ensures that patients receive the highest quality care possible. But how can we measure these patients' outcomes? And how do we measure this across different healthcare institutions?
Tiro.health collaborates with the Flemish Hospital Network KU Leuven and medical staff from 30 hospitals to develop multicentric clinical benchmarks reporting on patient outcomes. We mapped data from various medical specialties using the standardized terminology SNOMED CT and the data exchange standard FHIR. Clinicians participating in the project can access the dashboards at any time, enabling them to view both their own data and the aggregated data from other hospitals that have been benchmarked against theirs. One of the delivered projects is the quality dashboard for the “Flanders Inguinal and femoral hernia Prospective Registry” (FLIPR). Part of the data for this project is used for scientific research. Another presentation of the results will be shown at the Belgian Surgical Week 2023 ( “A 5-year analysis of groin hernia surgery in Flanders with a focus on laparoscopic groin hernia surgery”)
Let’s kick in some open doors. The main challenge for multicentric data analysis is the quality of the data. Hospital networks have already taken big steps in proper data collection these past years by introducing structured electronic forms in a wide range of pathologies.
There are however still some problems. There are differences across hospitals in the use of these forms and even within a hospital each clinician has his/her own way of completing the forms. Furthermore, the data is not captured according to the latest standards. In most cases, the data export is a CSV file which can contain over >1000 columns with self-chosen arbitrary terminology. Surgical procedures or diseases could have different names depending on which form they came from. Feel the headache already?
For the FLIPR project, the data is luckily of high quality. However, we must keep in mind that we are dealing with "real world data", which will have all sorts of shortcomings. Therefore, various assumptions are made when reporting on clinical outcomes. For example, for the follow-up after abdominal surgery, various measurements such as pain scores are taken at different time points (up until 5 years after surgery). There are some assumptions to be made such as setting time frames for each time point and deciding how to aggregate in case of multiple scores. In either case, it is important to clearly document the chosen approach and the underlying assumptions to ensure transparency and validity of the data analysis.
In any benchmark project, we will need a common data standard to map the data to. Otherwise it can be difficult or impossible for different systems to exchange and use data effectively with custom mappings for each data model. This can lead to inefficiencies and a lack of coordination in care, which can have negative impacts on patient outcomes.
Enter FHIR. FHIR is an interoperability standard for exchanging healthcare information electronically. The benchmarking applications that display quality indicators are fed by FHIR resources, which enables the calculation of quality indicators independent of the source of the data. We use this FHIR standard to standardize the exchange of medical data. A FHIR resource contains information about a specific concept within healthcare, such as a patient, operation, observation, medication, etc. By using the same international standard, we can ensure that we register the same medical data and guarantee the quality and format of the data.
For the project, FHIR profiles are created to validate resources. FHIR profiles will allow you to specify a set of constraints and/or extensions to the FHIR resources. They describe the expected output/resource, including example resources to clarify the profiles. The use of profiles ensures that the data captured in the resource is consistent and complete, which is essential for reliable multicentric benchmarking. FHIR profiles can be shared across institutions. One way is to publish them on simplify.net.
Example of FHIR procedure:
As you can see a set of medical codes are used to describe the procedure. Since there are various procedures possible for an hernia repair a set of included (SNOMED CT) codes will be defined. Any of the procedures codes used for this studies needs to comply with one of these codes or a child concept of the codes:
Multicentric data analysis and benchmarking not only improves the quality of care provided but also enhances the accuracy of medical reports. By demonstrating the importance of qualitative data through these dashboards, clinicians are encouraged to adopt a more structured reporting approach with a higher level of completeness. This not only benefits the benchmarking project but could also have a significant financial impact because of the improved accuracy of the medical reports. Furthermore, the multicentric standardized data generated by the platform can aid in scientific research, making it valuable for the medical community.
The bigger picture
At Tiro.health we believe in smarter medical notes. Our clinical report tool Atticus contains reusable templates designed for specific care paths in close collaboration with clinical experts. The completeness of the reports is checked simultaneously in order to comply with the agreed FHIR profile or other data targets. Upon completion, the reports can easily be exported in a standardized format (FHIR, OMOP CDM) and shared within and across healthcare organizations and researchers. The clinical dashboards are automatically updated when new reports are available. Our terminology server and auto-complete algorithm speeds up the note-taking process while ensuring standardized medical concepts (SNOMED, ICD10, LOINC,…).
Are you looking to improve the quality of your healthcare data and enhance collaboration between other healthcare organizations and researchers? Don’t hesitate to contact us!