Development and Application of a Human Reliability Analysis Method for Radiotherapy Applications

Abstract

Radiotherapy treatment is a complex process that involves communication between multiple expertise and continuous interaction with human-machine interface systems, to ensure safe and efficient patient handling. These attributes present a risk of failure with the consequence that patient safety is compromised, and incidents or accidents may occur. A common approach to prevent such risks is by reducing undesired occurrences. This is achieved by conducting retrospective analysis of accidents and incidents aimed to identify and classify the contributing factors, and, then, recommend prevention/mitigation strategies in form of directives and guidelines for patient safety for the clinics. These form the basis for the quality assurance program in each clinic. Literature research of outcomes of retrospective analyses of incidents and accidents from global databases, literature, and reports, indicate humans to be dominant contributors in 82-97% incidents. Recent safety guidelines point to the need of proactive risk assessment, building on and advancing beyond retrospective investigations. For this purpose, Failure Mode and Effects Analysis and Probabilistic Safety Assessment studies are conducted and have produced useful results; yet, when adopting these techniques, the systematic inclusion of possible human failures in the safety assessment is challenged by the lack of methods directly applicable to the specific radiotherapy domain. Indeed, as shown by literature research, Human Reliability Analysis (HRA) methods have been evaluated for their applicability to radiotherapy or healthcare in general to tackle and model human failures. Their application to these domains revealed that the existing methods do not address several human tasks specific to healthcare nor do they address the specificities of the radiotherapy context. The need to tailor HRA methods to specific domains is further supported by recent developments of HRA methods addressing domain-specific tasks and error producing conditions, e.g. railways, nuclear etc. Therefore, this thesis develops the first HRA method for radiotherapy domain and applies the method to study failure sequences in the radiotherapy workflow of a specific therapy center, the Center for Proton Therapy at the Paul Scherrer Institute of Switzerland. First, the thesis identifies and characterizes the taxonomies of factors influencing the radiotherapy personnel performance (performance influencing factors (PIF)) and tasks representative (Generic Task Types (GTTs)) of the radiotherapy domain that formed the building blocks of the HRA method. A total of six GTT and nine PIFs with definitions are developed for the method. A generic methodology is proposed to systematically and traceably identify set of PIFs affecting a GTT. It includes direct use of a cognitive framework to progressively map GTTs to failure modes, failure causes, failure mechanisms and PIFs. This provides a strong theoretical basis to the method. Then, the methodology is applied to the radiotherapy domain and develops GTT-PIF structures for the method. A total of eighteen GTT-PIF structures are developed for radiotherapy based on the proposed methodology. Further, these structures are validated against existing literature. Building on the developed qualitative assessment, the thesis addresses the quantification approach for the developed HRA method. To this aim, the Decision Tree (DT) methodology is chosen as the quantification methodology to compute the influence of the identified PIFs on the failure probabilities of the GTT-failure mode. Eighteen DTs are developed (one for each GTT-failure mode- PIF structure), in which (a) each branch point is the PIF and (b) each DT path represents the Human Error Probabilities (HEPs) due to the influence of a PIF or of a combination of PIFs. Once developed, the HEPs are estimated for paths of the DTs via a structured elicitation of judgment from domain experts. The experts assess the importance of specific human factors on the failure probability by means of a qualitative scale. Expert inputs are converted into statements about the order of magnitude of the probability values; these statements are then combined via an expert aggregation method, developed specifically for HRA. To build confidence on the developed methodology, the thesis validates the elicitation results against relevant applicable HEPs from existing HRA methods. Finally, the thesis combines the results of the two building blocks, i.e. the identified GTT-PIF structures and DT for the quantification of the HEPs and investigates ten failure sequences for the 4D radiotherapy treatment workflow at Paul Scherrer Institut, to systematically assess and quantify the associated failure probabilities. The analysis transferred into safety-enhancing proposals related to the implementation of checks and to the improvement of their effectiveness.