Advanced Pervasive Clinical Decision Support for the Intensive Care Unit

Abstract

Clinical decision support systems (CDSS) already have proven their use in supporting physicians and nurses in their daily responsibilities. However the uptake of these systems remains low. This is due to a number of reasons, such as little to no integration within the workflow of the medical staff, a multitude of available devices and a large amount of medical data that has to be processed. This research focusses on closing the communication gap between software engineers and domain experts and on integrating CDSS into the workflow of the medical staff by providing support at the right time, at the right place on the most suitable device, based on the current context. This way the variation in care can be reduced and costs are minimized. I. BACKGROUND TO THE RESEARCH As the complexity and size of medical information and data keeps rising, the need for clinical decision support systems (CDSS) grows at an ever increasing pace. CDSS are computerdriven technology solutions, developed to provide support to physicians, nurses and patients using medical knowledge and patient-specific information. Thus, these systems will not replace the medical staff, but will merely give advice and guidance. This way, they are able to take all relevant data and information into account. Examples of such data are information from databases or the electronic health record (EHR) [1]. By filtering the information in an intelligent manner and presenting it to the medical staff at the appropriate moment, these systems can improve health care [2]. CDSS can be used in every aspect of the care process, from preventive care and diagnosis to monitoring and follow up. Studies have already shown that these systems improve quality, safety and effectiveness of medical decisions, resulting in improved patient care, higher performance of the medical staff and more effective clinical services [3]. Nevertheless, the uptake of CDSS is rather low and this is due to a number of factors [4]. First, one of the main problems in the use of CDSS is the integration of applications into the current workflow of the medical staff [5]. Kawamoto et al [6] concluded that CDSS are more successful when integrated into the work process of the medical staff. This also means the integration with existing information systems and devices in the hospital. In a computerized Intensive Care Unit (ICU), a computer is located next to every bed. In addition, each department also has a unit PC and physicians usually have a personal desktop and telephone. An overview of this situation is shown in Figure 1. Moreover, the number of devices on an ICU is steadily increasing in recent years [7]. Currently, these devices are not optimal embedded within CDSS. Sharing information at the right time and place has a large influence on the use of these systems and on the performance of the medical staff, moreover it is time-saving [8]. Second, next to offering information to the users at the right time and place, the context of the user is also an interesting point of focus. Context-aware applications can take technologies, such as wireless technologies, sensors, mobile tools and handheld computers or smartphones into account. This can support health care providers in managing their tasks and will at the same time increase the quality of care [9]. In particular, ICUs sometimes contain complex and difficult situations and pose an interesting challenge for these systems. The third problem is representing all the relevant information of a specific patient. In the ICU, up to 200,000 parameters are collected for each patient on a daily basis [10]. These parameters are mainly originating from examinations and from monitoring data. Visualizing this data in an optimal way and selecting only the most relevant information is a challenging task [11]. It is also important to build new knowledge from the gathered data [12]. Last, CDSS are often based on clinical guidelines. Clinical guidelines are systematically developed documents used in hospital settings to standardize and structure treatment or diagnostic processes [13]. A typical guideline departs from observations, employs evidence-based practices and hence leads to conclusions and recommendations. If necessary, it can be deviated from, based on specific hospital policies or expert knowledge. Thus, guidelines are an indispensable tool in order to guarantee a high quality of care, while restricting redundant expenses [3]. Currently however, they are largely handwritten, text-based or depicted as flowcharts. The translation of written documents into working applications is very time-consuming and requires software developers to be thoroughly familiar with the medical problem domain. It requires a strong cooperation between domain experts and software developers, which is complicated by the communication gap between both worlds [14]. Even more important, multiple representation languages