Dealing with uncertainty

Abstract

Processes in the healthcare domain are characterized by coarsely predefined recurring procedures that are flexibly adapted by the personnel to suite-specific situations. In this setting, a workflow management system that gives guidance and documents the personnel's actions can lead to a higher quality of care, fewer mistakes, and higher efficiency. However, most existing workflow management systems enforce rigid inflexible workflows and rely on direct manual input. Both are inadequate for healthcare processes. In particular, direct manual input is not possible in most cases since (1) it would distract the personnel even in critical situations and (2) it would violate fundamental hygiene principles by requiring disinfected doctors and nurses to touch input devices. The solution could be activity recognition systems that use sensor data (e.g., audio and acceleration data) to infer the current activities by the personnel and provide input to a workflow (e.g., informing it that a certain activity is finished now). However, state-of-the-art activity recognition technologies have difficulties in providing reliable information. We describe a comprehensive framework tailored for flexible human-centric healthcare processes that improves the reliability of activity recognition data. We present a set of mechanisms that exploit the application knowledge encoded in workflows in order to reduce the uncertainty of this data, thus enabling unobtrusive robust healthcare workflows. We evaluate our work based on a real-world case study and show that the robustness of unobtrusive healthcare workflows can be increased to an absolute value of up to 91% (compared to only 12% with a classical workflow system). This is a major breakthrough that paves the way towards future IT-enabled healthcare systems.