Mapping Patient Data to Colorectal Cancer Clinical Algorithms for Personalized Guideline-Based Treatment.

BACKGROUND

Colorectal cancer is the most commonly occurring cancer in Germany, and the second and third most commonly diagnosed cancer in women and men, respectively. In this context, evidence-based guidelines positively impact the quality of treatment processes for cancer patients. However, evidence of their impact on real-world patient care remains unclear. To ensure the success of clinical guidelines, a fast and clear provision of knowledge at the point of care is essential.

OBJECTIVES

The objectives of this study are to model machine-readable clinical algorithms for colon carcinoma and rectal carcinoma annotated by Unified Medical Language System (UMLS) based on clinical guidelines and the development of an open-source workflow system for mapping clinical algorithms with patient-specific information to identify patient's position on the treatment algorithm for guideline-based therapy recommendations.

METHODS

This study qualitatively assesses the therapy decision of clinical algorithms as part of a clinical pathway. The solution uses rule-based clinical algorithms, which were developed based on the corresponding guidelines. These algorithms are executed on a newly developed open-source workflow system and are visualized at the point of care. The aim of this approach is to create clinical algorithms based on an established business process standard, the Business Process Model and Notation (BPMN), which is annotated by UMLS terminologies. The gold standard for the validation process was set by manual extraction of clinical datasets from 86 rectal cancer patients and 89 colon cancer patients.

RESULTS

Using this approach, the algorithm achieved a precision value of 87.64% for colon cancer and 84.70% for rectal cancer with recall values of 87.64 and 83.72%, respectively.

CONCLUSION

The results indicate that the automatic positioning of a patient on the decision pathway is possible with tumor stages that have a less complex clinical algorithm with fewer decision points reaching a higher accuracy than complex stages.