Flexible pathway orchestration engine for healthcare using BPMN and workflow systems

Abstract

Healthcare is facing rising costs; there is pressure from insurances and governments to optimize the care delivery in order to reduce costs. However, creating software for supporting healthcare processes is challenging. Clinical processes are knowledge-intensive and highly regulated. This means it cannot be automated so easily. In addition to that, the uniqueness of each patient and the different type of care that can be delivered (planned or emergency) require flexibility. Despite those complications, the industry needs formalized processes and software that could support those. Healthcare workers could increase their efficiency and performance if they could easily track, oversee and get suggestions on what are the recommended actions for every patient case. Based on previous research we found out that checklists are a recognized way to interact with the user and support healthcare processes. The idea was to implement the checklist into a workflow engine. The problem was that workflow systems are too rigid and do not allow deviations. In this research, we present a prototype of a flexible engine that works with checklist models. We provide a generic modeling approach for modeling processes using checklists: a series of stages executed according to a progressive order and a set of items considered the checks on a checklist grouped in specific stages. Users are able with a sample user interface to mark the checks in the checklist and proceed with the checklist. They are also allowed to deviate from the desired path (order of stages) and change the embedded model during runtime. We identify three operations that would require deviations:

1. Allow postponing of a set of tasks and allow the process to proceed. 2. Allow the delegation of a set of tasks to the responsible for the next stage of the process. Like the previous operation, the delegation allows the process to proceed. 3. Allow the user to do task meant to future stages in the process and not yet enabled.

Those three deviations are the ones implemented in the prototype. In addition to those we have evaluated other type of deviation (like rollback), but hasn’t been implemented. For changing operations, we have implemented two different actions in case the user wants to change checks of the stage currently active or future ones. We have prevented the user to make changes on items in past stages (unless they are still to be completed). All those deviations were used to explore several circumstances where forgiveness was key to support user deviation. Having a reusable model also allowed us to test in different scenarios and connecting with other home-grown systems. The goal of the prototype is not only to support users with their process but also to record the timing, deviations, and changes in the process. We also identify some features that could be part of future work such as further metrics, optimization algorithms and insight on how to improve the running process. Based on our sessions with potential stakeholders we found out that forgiveness is a desirable characteristic in complex healthcare software and the models created to provide a flexible solution.