| Popis: |
This thesis investigates requirements engineering methods based on process modelling for Electronic Healthcare Record (EHR) systems. The relation between software requirements and user workflows is essential in healthcare settings: EHRs are expected to improve clinical and administrative workflows. In turn, the new workflows are expected to satisfy a number of business goals. If a new software system does not support the desired clinical workflows or patient journeys, then its value and benefits are often disputed by stakeholders. Our hypothesis is that requirements engineering methods based on process models will contribute to the overall success of EHR projects in the industry. By success, we mean software systems that are in use and meet the business benefits expected of them. The experiments presented in this thesis are aimed to develop and evaluate a method that allows business analysts to make use of process models during requirements engineering for EHRs. The goal of the method is to ensure the software specification is aligned to and supports the user workflows. Each of the four experiments addresses a specific research objective, and thus the findings from each experiment constitute the basis for one of our four contributions to science. / Experiment 1: Relating Goal Oriented Requirements Engineering and Process Modelling: This experiment investigates the design of a common framework for describing process models and software requirements. It relates the KAOS framework for goal oriented requirements engineering and the Business Process Modelling Notation (BPMN). Our goal is to facilitate requirements elicitation. Specifically, business analysts using our framework should be able to reason about the alignment of the software specification to the business processes, and identify specific changes that improve this alignment (either changes in the design of the system, or changes in the business processes). This first experiment was conducted as part of the WellbeingUCL project, supported by Boots. / Experiment 2: Inferring Goal Models from Process Models: The second experiment investigates a method for business analysts to derive software requirements from process models. The purpose for defining such a method is to provide sufficient guidance to business analysts, during requirements elicitation. Our aim is to help business analysts elicit meaningful goal models and shape the design of the system-to-be, in light of these goals. A number of heuristics to facilitate requirements elicitation are proposed and evaluated, considering the trade-offs between a fully automated and a human driven process. / Experiment 3: Electronic Healthcare Record for Bupa: The third experiment evaluates the requirements engineering method during an EHR implementation for a chronic condition management service delivered by Bupa nurses in South West England. Action research is used to assess the impact and fit of the requirements elicitation process, in relation to the current work practices of business analysts in the industry. The extended KAOS framework and goal inference heuristics have been used to inform the final software specification, guide the workflow redesign and clarify the business benefits. From a project management perspective, this experiment evaluates how the KAOS method aligns with the Agile and Lean methodologies used in Bupa. The project has delivered an EHR system actively used to support the care of 2,600 patients. / Experiment 4: Personal Health Record for Nuffield Health: The fourth experiment evaluates the extended KAOS framework when developing a new digital customer proposition with an underlying EHR system. It investigates how consumer journeys can be modelled as KAOS process models. Of specific interest is the ability of the framework to clarify the responsibility assignments among the different agents (i.e. system components) that need to collaborate to deliver the end to end customer journey. The experiment was run as an action research project, in partnership with Nuffield Health. The results have informed the architecture of an open source personal health record for lifestyle data. / Contributions to science: This thesis advances the field of requirements engineering by introducing and evaluating a requirements elicitation method based on business process models. It also presents new evidence into the use of goal oriented requirements engineering for the design and implementation of EHR systems in the industry. Our four contributions to science directly follow from the results of the four experiments conducted as part of this research. Our first two contributions cover the conceptual framework and our proposed method for requirement elicitation based on process models. Our last two contributions present evidence for the practical use and benefits of our goal oriented requirements engineering method in industry based projects. First, we present an extension of the KAOS requirements engineering framework which includes a business process view with clearly defined syntax and execution semantic. This approach ensures process models and goal models have a shared semantic. A new concept, that of Intentional Fragment, captures the explicit relation between fragments of a process model and a specific goal. We also define additional consistency rules, to clarify how the process view relates with other KAOS models: object, agent and operation model. Secondly, we present a set of goal inference techniques to help analysts build goal models starting from process models. In effect, analysts can start from the artefacts that are most familiar to them (i.e. the workflow models) and gradually derive a goal model for the system-to-be. A set of 12 heuristics have been fully defined and integrated into a semi-structured method for goal elicitation. Our third contribution is an evaluation of how the goal oriented requirements engineering method (incorporating workflow analysis) supports the design and deployment of a EHR system in a clinical setting. The project was representative for the challenges faced by healthcare organisations wishing to deploy EHRs: quality of care standards that impose constraints on process redesign; legacy systems that have shaped the workflow; organisational complexity and competing stakeholder interests. We show that by methodically applying our goal inference techniques we were able to produce a valid goal model starting from models of the nurses workflows. The resulting goal model was used to reason about alternative design options in the system-to-be, and to clarify the benefit case in deploying the EHR system. Fourth, we examine the requirements engineering process for an EHR system meant to support a new customer proposition. This project was representative for the challenges faced in the digital health industry: a target consumer journey driven by user experience research; many different systems required to collaborate; focus on the architectural design of the system. We show that we can apply our goal inference techniques to customer journey maps and produce a meaningful goal model. This has been used to shape the architecture of the EHR system and reason about integration requirements. We also argue that our goal inference techniques complement agile development practices used within the organisation. |
