Quality standards in radiation medicine

Identifying and conducting "best practice" medicine is arguably the ubiquitous goal of practitioners. However, to distill the many available quality standards, guidelines, recommendations, and indicators down to a best practice set requires a logical schema to group standards addressing similar quality issues and, from manageable lists of related standards, to extract the essential dimensions of quality. The purpose of this study was to explore a method of collating publicly available quality standards, in this case in radiation therapy, using a 2-step decision tree approach with statistical analysis. Successful grouping into manageable lists, addressing related quality issues, informs the ongoing development of quality indicators that are one expression of "best practice."

A comprehensive literature search was used to identify quality standards currently in use and publicly available. Using 2 decision trees, 5 evaluators assigned each standard to Donabedian's structure, process, or outcome and also to the target of the standard: patients, staff, equipment or clinical process, or organization for a total of 3 × 4 = 12 primary categories.

A total of 454 radiation medicine program quality standards spread across 8 national and international documents was identified. Agreement between the 5 evaluators, using the free marginal kappa statistic, ranged from fair to almost perfect. In all but 2% of 5 × 454 evaluations were the evaluators able to assign a statement to categories in the decision trees suggesting that these trees are appropriate to the task. In only 3/454 was a majority (≥ 3/5) decision not reached on the assignment to structure, process, or outcome. Sixty-four percent of the standards were identified with structure, 26% with process and 10% with outcome.

Donabedian's model constitutes a reliable method of managing quality standards. The 2-step decision tree framework can be applied to inform the further development of national and international quality standards.