Article Text

Download PDFPDF

2022-RA-1085-ESGO Using data-driven algorithms and real-world data for updating endometrial clinical practice guidelines
  1. Olga L van der Hel1,
  2. Kees Ebben1,
  3. Thijs van Vegchel1,
  4. Arturo Moncada Torres1,
  5. Jurrian van der Werf1,
  6. Maaike A van der Aa1 and
  7. Cor D de Kroon2
  1. 1Research and Development, IKNL, Utrecht, Netherlands
  2. 2Gynecology, Leiden University Medical Center, Leiden, Netherlands


Introduction/Background Clinical practice guidelines (CPGs) are commonly structured as manuals, where best practices are described as free text. Since most oncological CPGs have an extensive care pathway, keeping these CPGs unambiguous and up-to-date is complex. We propose an innovative approach that allows guideline developers to take action and consider updates when scientific developments of CPGs (represented by the National Comprehensive Cancer Network, NCCN) or notable trends in clinical practice (represented by the Netherlands Cancer Registry, NCR) are identified.

Methodology First, the Dutch national and NCCN endometrial cancer CPGs were translated into clinical decision trees (CDTs). Then, we requested an endometrial cancer dataset from the NCR and mapped it onto the CDTs. Thereafter, we designed an information standard by applying FAIR principles. Finally, analysis and comparison functionalities were made available in a prototype dashboard. Predetermined thresholds were implemented that raise a notification to the guideline developers when numbers are outside expected range.

Results Both CPGs were successfully translated into CDTs. This yielded 10 and 15 CDTs, 58 and 72 data-items (patient and disease characteristics), 57 and 97 subpopulations, and 61 and 138 recommendations for the Dutch and NCCN CPG, respectively. Also, the NCR dataset was successfully mapped onto 5 CDTs from the Dutch CPG. The data were projected onto the CDTs. We identified adherence levels for all subpopulations and alternative treatments for non-adherent cases.

Conclusion Applying our method in a dashboard identified ambiguous, redundant, and incomplete sections of the Dutch CPG for endometrial cancer and raised notifications for relevant observations. This data-driven approach could serve as automated surveillance to determine best clinical practice for patient (sub)populations and accelerate the creation of living recommendations. Moreover, this approach is suitable for applications in other diseases and settings.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.