CareVis: Integrated visualization of computerized protocols and temporal patient data
Introduction
Visualization is an integral part of intelligent data analysis. On the one hand, visualization is applied for presenting results and on the other hand, human perception might be utilized for driving the analysis process. In the medical domain, mostly patient data measurements are used as basis for analysis either in form of high-frequency data of intensive care settings or low-frequency data, e.g., for long term studies. Current visualization methods are mostly bound to the representation of such measured patient data only, which can be subsumed by the term “data visualization”.
But, there is much more information to be taken into consideration in the analysis process. One of these informational pieces is treatment information, that is, data on which therapeutic steps have been taken at what time, for how long, how often, and the like. So far, contextual information on treatment steps and performed treatments is mostly excluded from first-hand data analysis. The integration is either only performed mentally by physicians or worse, contextual information is lost completely. Such information could be an important source for finding reasons and explanations for certain phenomena in the measured patient data. The goal of this work is the integration and combination of various kinds of data as well as information and presenting it in a coherent way for supporting the data analysis process.
Computer-supported protocol-based care is a field of research that aims for supporting the treatment process along protocols semi-automatically by the use of information technology. The core entity, medical treatment plans, are complex documents, mostly in the form of prose text including tables and figures [1]. Protocol-based care utilizes clinical protocols to assist in quality improvement and reduce process irregularities. Such clinical protocols are a standard set of tasks that define precisely how different classes of patients should be managed or treated. They can be seen as reusable definitions of a particular care process. Not much work has been done in order to communicate the computerized treatment plans to the medical staff. Moreover, a combination with the presentation of patient data when treating a patient along a plan for monitoring and analytic tasks has been considered even less. However, the integrated visualization of patient data and medical treatment plans could be of great assistance to ease the complex task of analyzing medical data and protocols.
In the upcoming section, we will provide a task and data analysis of the problem domain followed by a discussion of related work in Section 3. After this, we will describe the design of our visualization approach CareVis in Section 4. Subsequently, we discuss our user-centered design methodology including prototype implementation, evaluation, and user study results in Section 5. Finally, we describe how users’ tasks are supported and sum up our findings.
Section snippets
Problem analysis
In this section, we analyze the problem domain from a data as well as user’s perspective. At the beginning, we will shortly explain the plan representation language Asbru that our project is based upon. Following this, we illustrate use-cases in form of scenarios and identify major task classes. Finally, we will combine the constraints and requirements to a summary of data characteristics that need to be dealt with.
Related work
In the following, we discuss related work in the areas of medical treatment planning, information visualization, and commercial medical software.
Design of the integrated visualization CareVis
The underlying data structure that we want to communicate to medical domain experts is very complex. Since none of the existing visualization methods can be used to represent all needed data characteristics, we decided to use the approach of multiple views [24]. Multiple views are a well known information visualization method, whereby a number of representations that focus on different aspects of the data are provided for a common underlying data structure [25].
User-centered design, prototype, and evaluation
“New medical information systems, no matter how fast, inexpensive, and easy to use, will not be used more widely until it has been demonstrated to practitioners that these systems provide answers that help solve the problems of patient care.” [31]
When developing our interactive visualization methods, we put forward a user-centered design approach. This included a user study, the discussion of the designed methods in a review step, and the evaluation of our Java prototype as described in the
Supporting users’ tasks
So how can Markus Zolte, Andrea Habacher, and Heinrich Kovanic benefit from our visualization methods in accomplishing their work tasks as described in Section 2.2?
Conclusion
Our goal was to develop visualization and interaction methods for supporting medical personnel in computerized protocol-based care. To achieve this goal, we had to consider several data aspects like the logic, structure, and temporal constraints of plans as given at design-time, data of instantiated plans at execution-time, as well as patient data in form of parameters and variables. Applying a multiple views approach helped to master the complexity of the underlying data structure while using
Acknowledgment
This project is supported by “Fonds zur Förderung der wissenschaftlichen Forschung—FWF” (Austrian Science Fund), grant P15467-INF.
References (34)
- et al.
Approaches for creating computer-interpretable guidelines that facilitate decision support
Artif Intell Med
(2004) - et al.
Visualizing queries on databases of temporal histories: new metaphors and their evaluation
Data Knowledge Eng
(2003) - et al.
The visual display of temporal Information
Artif Intell Med
(1991) - et al.
Guidelines for clinical practice: from development to use. Institute of Medicine
(1992) Plan management in the medical domain
AI Commun
(1999)Use of algorithms in clinical practice guideline development: methodology perspectives
Clin Practice Guideline Dev: Methodol Perspect
(1995)- Society for Medical Decision Making. Proposal for Clinical Algorithm Standards. Med Dec Mak...
- et al.
Graphical summary of patient status
Lancet
(1994) - et al.
Visualizing the logic of a clinical guideline: a case study in childhood immunization
Methods Inf Med
(1997) - et al.
LifeLines: using visualization to enhance navigation and analysis of patient records
Stefanelli, comparing computer-interpretable guideline models: a case-study approach
J Am Med Inform Assoc (JAMIA)
Metaphors of movement—a user interface for manipualting time-oriented, skeletal plans
Artif Intell Med
Visualizing temporal clinical data on the WWW
Readings in information visualization: using vision to think
Cited by (64)
A data science approach to drug safety: Semantic and visual mining of adverse drug events from clinical trials of pain treatments
2021, Artificial Intelligence in MedicineSimulation of patient flow in multiple healthcare units using process and data mining techniques for model identification
2018, Journal of Biomedical InformaticsSpatiotemporal data visualisation for homecare monitoring of elderly people
2015, Artificial Intelligence in MedicineCitation Excerpt :In timelines, events are represented as geometric figures, arranged horizontally according to the time they occurred and displayed in one or multiple rows depending on their categories, as in Gantt charts [4–6]. In the medical domain, the visualisation of patient records can be done by events over the timeline as in LifeLines [11], KHOSPAD [12], KNAVEII[13], Care Vis [14], LifeFlow [15], and OutFlow [16]. However, in AAL systems, sensors register a large number of events and basic timeline-based representations are not effective.
Public health nurse perceptions of Omaha System data visualization
2015, International Journal of Medical InformaticsInteractive visual guidance for automated stereotactic radiosurgery treatment planning
2015, Expert Systems with ApplicationsCitation Excerpt :For this reason, successfully integrated decision support systems should provide a well-balanced allocation of tasks between experts and automation, motivating the interface-based class of extensions. Although interfaces are a well-established method to support human decision making (Grandjean & Kroemer, 1997; Korhonen & Wallenius, 1988; Lotov, Bushenkov, & Kamenev, 2004; Wickens et al., 2004), and have been more specifically addressed as useful in the field of medicine (Thyvalikakath et al., 2014; Aigner & Miksch, 2006; Gschwandtner, Aigner, Kaiser, Miksch, & Seyfang, 2011), they are still under-utilized in public health applications (Aigner & Miksch, 2006; Thyvalikakath et al., 2014; Yasnoff & Miller, 2014). Interfaces that are implemented frequently lack sophistication, making them less effective in their task of reducing the user’s cognitive load (Wickens et al., 2004; Yasnoff & Miller, 2014).
ClinicalPath: A Visualization Tool to Improve the Evaluation of Electronic Health Records in Clinical Decision-Making
2023, IEEE Transactions on Visualization and Computer Graphics