TimeSpan is a visual exploratory tool designed to gain a better understanding of the temporal aspects of the stroke treatment process. Working with clinical stroke experts, we seek to improve outcomes for stroke victims. Time is of critical importance in the treatment of acute ischemic stroke patients. Every minute that the artery stays blocked, an estimated 1.9 million neurons die. Consequently, there is a critical need for efficiency of stroke treatment processes. Optimizing time to treatment requires a deep understanding of interval times. Stroke health care professionals must analyze the impact of procedures, events, and patient attributes on time---ultimately, to save lives and improve quality of life after stroke. Based on experts' requirements, we designed TimeSpan, a web-based tool for exploring multi-dimensional and temporal stroke data. The proposed system incorporates factors from stacked bar graphs, line charts, histograms, and a matrix visualization to create an interactive hybrid view of temporal data.

Bibtex

@article{Hosseinkhani:2015:TimeSpan,
Author = {Mona Hosseinkhani Loorak and Charles Perin and Noreen Kamal and Michael Hill and Sheealgh Carpendale},
Journal = {TVCG},
Title ={TimeSpan: Using Visualization to Explore Temporal Multidomentsional Data of Stroke Patients},
Volume = {To appear},
Year = {2015},
project = {TimeSpan},
}

Slides

Extended Materials

Focus Group Insights

In this section, we provide examples of how TimeSpan visualizations led to insights in the focus group discussion.

1. Simplicity to Complexity

We moved our explanation in the focus group session from simple to complex interactions. This simplicity to complexity of interactions is demonstrated in the attached video to the submission which can also be found in the following.

TimeSpan: Using Visualization to Explore Temporal Multidomentsional Data of Stroke Patients from mona on Vimeo.

2. Treatment Process Order Change

During our field observation and interviews, we were told that the process of delivering tPA was always in the following sequence: symptom onset, EMS patch time, door time, CT scan, tPA, and endovascular therapy. Using TimeSpan, we found one patient that his/her timings were not in accordance with the above sequence of events. The time intervals for this patient are showing in Figure 1.
Figure 1: The treatment process for the highlighted patient does not follow the usual order. In other words, this patient receifigcaptionved endovascular therapy before administering tPA. That is the reason we see an overlap of the red bar over the green bar.

Initially, we considered that case as a data entry error. However, after looking through the visualization in the focus group session and visually recognizing that particular patient, our participants mentioned that the treatment process could change sometimes and those cases are in fact interesting for them. “The ordering may not always happen such that tPA will happen before groin puncture. Sometimes tPA is given in the angio suite after groin puncture as well.” Another participant said “In the visualization, it would just have to overlap the red and the green like there is one there.” This possibility of changing the order of events was not mentioned during the initial interviews, but looking at the visualized data brought this case to participants' minds.

3. Endovascular Therapy Based on Age

As part of the focus group, we illustrated the functionalities of TimeSpan by showing some data discoveries while we were exploring the data. For example, we had discovered that most of the patients who got endovascular therapy were among the younger patients while senior patients were less likely to receive it. This is demonstrated in Figure 2 where the baseline is set to tPA time and the patients are ordered according to their age.
Figure 2: The baseline is set to tPA time and the Patients are ordered according to their age from left to right. This figure shows that the frequency of endovascular treatment is higher among younger patients. This finding was confirmed with our stroke experts and they started looking for possible reasons in the focus group session.

Discussing this finding with our domain experts in the focus group, one participant explained that stroke neurologists might already know about this fact: “I think that just you guys [turning towards a stroke neurologist] would know that older people have more tortuous arteries. So, likely they would have less endovascular treatments than younger patients. I think typically you tend to be more aggressive with younger patients.” However, one of our stroke neurology experts explained that this fact is worth further investigation as tortuous arteries might not be a problem and they may actually be age-biased. “I think that you point out something, like it's an observation. I think we have certain hypothesis on why that is but I think it might be important to look into it because everybody has certain pre-conceived notions on why we are doing things. I am just saying like if they all had tortuous vessels and they cannot get in there maybe yeah that is the explanation but maybe we are just age biased and we are slower with older patients.” Another participant added the fact that this needs further analysis. “We can do a chart on it, look at the imaging of all those patients, and did they truly have tortuous vessels or not.” She also liked the fact that visualization allows them to see some correlations. “So, this is really nice to kind of look for the associations.”

4. Long Transfer Times

Our participants to the focus group noticed that some patients with longer patch to door times are having longer transfer times compared to the rest of the patients. Figure 3 shows one of the patients whose timings caught our participants’ attention in the focus group session. “The transfer times are generally really short so I'm guessing all of those are [Hospital X]. But look at this guy here, look at this transport time. You know that he is coming from [Hospital Y] or somewhere. There is no way that was from [Hospital X].”
Figure 3: The highlighted patient is having a long patch o door time compared to the rest of the patients.
Afterwards, the participants started thinking about specifying and identifying the category distace (e.g., zone1, zone2, and zone3 within the city) and having some radious zones be collected, integrated, and visualized using TimeSpan. "We should have radius zones. Like zone 1 is this and zone 2, and 3 and see how they look like."

5. Pros and Cons of Visualizing Individual's Data

One of the stroke neurologists mentioned that she wants to know how she has performed in terms of treatment time: “I want to know how I'm doing". She also said that she is interested in comparing her performance to other neurologists: "I want to have my median time compared to others.” These comments were followed by a long discussion between the experts about the pros and cons of visualizing individual professional results and comparing them. "The problem is you can see this huge barrier, there's so many contributing factors to delay that you can't really put it down to physician." "So, if you get all the horrible patients from that month that just happen to come in and it takes forever to get the TPA that doesn't mean that you are slower than the next guy." A participant also added the fact that aggregated data such as median are helpful in getting an overall understanding of how people are doing. "So, I think that's where the median is important and then you can compare and see if it's really bad."

6. Data Requirement

Participants visually noted the missing patient attribute values in the collected data. The HEDA clearly emphasized that the data had been inconsistently collected (e.g., all the cross signs in Figure 4 indicate a data item that is not available). “It seems that we have a lot of missing data.” "If there are things that are mostly just missing data, I don't know, you know maybe that's not a useful data or we will just have to start collecting it right."
Figure 4: The cross signs in the HEDA area represent not available data.

Based on this observation, participants discussed the kind of data attributes they currently collect as well as the way they collect the data. "We have a lot of missing data ... its a hand collected data."
By looking at HEDA in Figure 4, participants also started asking for collecting new data and visually incorporating them in HEDA. As an example, a participant found that high blood pressure and its treatment is more important than coumadin. "I see that Coumadin made it to the boxes. I think there's a finite thing that we can put on there but blood pressure is important to put on there or treatment for blood pressure." "I almost feel like maybe coumadin is almost less important than blood pressure." Another participant mentioned that they need to collect data as they have not collected it so far. "We don't have the data for blood pressure for now."

7. Corresponding Visual Elements with Numbers

A participant indicated that he would like to have access to some aggregated numerical data about the selected patients using the selection tools: “just to have even a percentage or something like that will help us, like percent of stat-strokes.” “I think people like that cause I think visually you're looking for a pattern and it has to correspond with a number.” Figure 5 shows where the participant was pointing to when he was requesting to add new statistical data in the visualization.
Figure 5: One of the participants pointed towards the upper area of the selection tools and requested to add more aggregated numerical data about the selected patients in those areas.

8. Visual Separators

Participants expressed their need for visual separators in the detailed view. In particular, these separators would be helpful to focus on a contiguous subset of patients. A typical example would be to create separations when a given patient attribute is chosen to reorder the patients, or use different background colors to separate groups of patients. “I could sort of see following it down. Where one starts and where one category ends. I wonder if we can also have some color differentiation, it would just help to better see the categories.” Figure 6 shows an example of the place where the separators were requested by our participants.
Figure 6: TPA location has 4 distinct values which are 112 room, Angio room, CT scan room, and Emergency Room (ER). When reordering the patients based on this factor, visual separators among groups of patients having different attribute value were requested by participants.

9. Recognizing the Huge Varience in DTN Times

We also found that by merely looking through the visualization and without interacting with it, our participants started asking on why there are so much variance in treating patients in different months. “We have a huge variance here (Pointing at months with huge treatment variance).” “yeah, a huge variance.” Example of months with big variances are shown in Figure 7.
Figure 7: September 2013 and June 2014 are among the months with huge variance in treating patients.
Then, the participants started wondering whether this is due to patient factors or due to system related factors. “one thing as we know that there's patient factors which really we can't control.” “Like, we had a case during the weekend where a woman was brought in and the EMS said oh she was last seen normal at 7am but when we were looking at her onset history, her time was much more closer to the time we treated her. So you could see the human factors.” Another participant started arguing that they might be due to system factors which need further investigation. “But there are all the systems factors where some of them you can visually see where it's taking a long time to get to the CT scanner, I personally would see that as a systems factor, right.”

System Design

In this section, we provide a review of our goals in designing TimeSpan and its designing process. In designing TimeSpan, our goal is to represent our temporal, multivariate and multi-typed data in a single holistic view. This seems to be a crucial factor in this context as the system's nature is exploratory and its purpose is to discover the factors that are introducing delays in the treatment process. As these factors and their relations with each other are not known beforehand, we are trying to concentrate as much data as possible within a single view and assuming that every single factor matters in exploring and understanding the data. In TimeSpan, we made our design decisions based on iterative paper prototyping of various alternatives and asking for experts' feedback to compare different possibilities. Through this iterative design, we learned about people's preferences, suggestions, and critics, which eventually helped us to realize the principal of "getting the right design". Figure 8 demonstrates some of the sketches we showed to domain experts and discussed about them for informing the design of TimeSpan.

Figure 8: A number of our sketches in designing TimeSpan.

Authors


Mona Hosseinkhani
Mona is a PhD candidate working with Dr. Sheelagh Carpendale in the Innovis Group (Interactions Lab, Department of Computer Science) at the University of Calgary. She is interested in visualizing complex data in novel ways to make the data more accessible to the viewer.

Charles Perin
Charles is a post doctorate researcher in the Innovis Group (Interactions Lab, Department of Computer Science) at the University of Calgary. He graduated in PhD in 2014 from the university of Paris Sud-XI in collaboration with INRIA.

Noreen Kamal
Noreen is a Postdoctoral Fellow at the Cumming School of Medicine, Department of Clinical Neurosciences at the University of Calgary.

Michael Hill
Michael is a Professor at the Departments of Clinical Neurosciences, Community Health Sciences, Medicine and Radiology at the University of Calgary. He is also Director of the Stroke Unit for the Calgary Stroke Program, Alberta Health Services.

Sheelagh Carpendale
Sheelagh is a Professor at the University of Calgary where she holds a Canada Research Chair: Information Visualization and an NSERC/iCORE/SMART Industrial Research Chair: Interactive Technologies.

Please spread the word if you find this work interesting