Identifying best practice treatment strategies by incorporating information from similar healthcare pathways

Project summary

We hypothesize that similar healthcare pathways can be identified from routine clinical data and small data situations with multiple distinct pathways and decreasing numbers of observed transitions per pathway over time. Using the healthcare pathways of a particular small group of patients as a starting point, we will develop approaches to quantify the similarity of the pathways of other patients to this group based on multistate models. We will quantify uncertainty and construct confidence intervals for these measures, and use them to formulate hypotheses and corresponding tests that allow to decide on the similarity of two healthcare pathways. Furthermore, new clustering methods for several multistate models will be developed, for identification of patient groups that represent best practice healthcare pathways.

Our methods

  • Knowledge-driven modeling
  • Local perspective

Principal investigator 1

Doctoral researcher position 1

(supervised by PI Binder N.)


Investigations on new measures of similarity, power characterization of the bootstrap approach in real life settings, and evaluation of clustering procedures.


  • Master’s degree or equivalent in mathematics, (bio-)statistics, or similar
  • Programming skills in, e.g., Julia, Python, or R
  • Ideally, expertise in stochastic process modeling
  • Strong interest in modeling routine clinical data

Principal investigator 2

Doctoral researcher position 2

(supervised by PI Dette; located in Bochum)


Proofs of the validity and accuracy of bootstrap confidence intervals and tests, and the development of a spectral clustering algorithm.


  • Master’s degree or equivalent in mathematics with a strong background in statistics or in statistics with a strong background in mathematics

Administrative Manager

Marc Schumacher

Institute of Medical Biometry and Statistics,
Faculty of Medicine and Medical Center –
University of Freiburg