A Computational Method for Systematic Comparison of Clinical Practice Guidelines

Clinical practice guidelines are essential in improving oncology healthcare by translating scientific evidence into practice. Evaluating and comparing guidelines for the same medical condition across organizations and countries aids in assessing their validity. However, inconsistencies in their format and the topics covered hamper large-scale efficient guideline comparisons. This study applies a technical format to guidelines from two regions and introduces a computational method for systematic guideline comparison.

For every decision-making moment identified in the Dutch and US guidelines for endometrial cancer clinical decision trees (CDTs) were developed. These were organized into coherent care pathways and classified according to an oncological reference model. Subsequently, calculations and content comparisons were performed on the revealed concepts at a semantic level. Lastly, recommendations were compared across the subpopulation(s) defined within each CDT.

The method identified similarities, relations and differences at various levels between both guidelines. Analyses yielded 17 CDTs, 33 population characteristics and 75 interventions for the Dutch guideline, while the corresponding figures for the US guideline were 20, 61 and 111, respectively. Within the CDTs, 11 matching subpopulations were identified, of which 2 had identical recommendations, 5 had related recommendations and 4 showed mismatched recommendations.

The current computational approach supports structured comparisons in research settings and allows for application in clinical practice via a developed dashboard prototype. Computational guideline representations reveal gaps, differences and similarities across multiple levels. The study underscores the potential for increased utility of computable guidelines, with the developed method also applicable to various other rule-based instruments.