Survey Round Two – Evaluating TB Report Designs

Are you involved in TB management (clinical, lab or public health) or research? We invite you to participate in our second report design survey! Click the button below to get started!

start_survey

Using data from the first TB Report Design Study, we developed various prototype report designs during a one day design sprint in collaboration with Dr. Tamara Munzner’s human computer interaction research team at UBC. This event resulted in eight prototype reports, of which multiple approaches to data representation were identified.

The next step involves taking these prototype designs and evaluating them based on end-user preferences. If you are involved in TB management or research, we invite you to complete our second online survey. This will be open for approximately two weeks starting on September 12th, 2016.

See our previous post for further information on the TB Report Design project and interim results.

Reflecting on infovis, or “why are we doing this”?

Jenn Gardy here, Gardylab decorative figurehead. Now that Amazing Resident Geoff has got us set up with this blog for sharing our public health information visualization-related work (which I’m calling phinfovis for short), I thought I’d chime in with a little history and reflection on how we ended up here.

My first exposure to infovis came as a postdoc – I’d just been recruited to Bob Hancock’s lab at UBC, where together with my PhD supervisor Fiona Brinkman, the team was working on InnateDB – a massive database of all the biological interactions associated with the innate immune system. Part of InnateDB is an analysis platform that lets users upload gene expression data and look for over-represented GO terms and pathways, and the group was interested in what that platform might look like – how do we show complex interaction networks painted with gene expression data from multiple conditions and/or time series in a meaningful, intuitive, and interpretable way. That was what I was tasked with.

I searched the UBC faculty directory to find people who were working on visualization and human-computer interaction, and found Ron Rensink, who was working on things like change blindness (how, when confronted with two subtly different images, it takes us ages to spot the difference between the two – try it for yourself here). This is exactly the sort of thing you;d be dealing with as a biologist with two or more differently-painted networks, desperately flicking back and forth between the images, looking for some sort of interesting biology.

It turned out that Ron was on sabbatical that year, but he did point me to Tamara Munzner, a comp sci prof who had joined UBC a few years before. She had already worked with a few biologists on projects like sequence and tree comparison and was excited to try out the biological network space, having done her graduate work on massive graph visualization. And thus began a beautiful collaboration that is still active a decade later!

With Tamara and her graduate student Aaron Barsky, we created Cerebral, a Cytoscape plugin for laying out graphs according to subcellular localization (thereby making a graph into something that resembles the signalling pathway diagrams biologists know and love) and for easily visualizing multiple expression datasets painted atop said graphs.  Not only was it a cool contribution to the infovis community, but that little workhorse helped me analyse a load of immunology datasets. Mind you, I still can’t tell you whether IL1B is pro- or anti-inflammatory. I really don’t like immunology.

Anyway, after the Hancock Lab postdoc I joined BCCDC, where I began working on using genomics for reconstructing infectious disease outbreaks (you can read all about that on my other lab blog). Inspired by my work on Cerebral and the AMAZING data and vis work done by one of my best friends, Jer Thorp, I always tried to inject a bit of vis sensibility into the research outputs I was creating. A couple of examples I liked included this:

nejmnetwork
This figure combines a two-clade phylogeny (lineage A and B) with a social network (circular nodes, edges) and some clinical information (node shading) to show the path of TB in an outbreak.

And this:

This reconstruction of a measles outbreak combines a contact network (lines) with an epidemic curve-style display (x-axis), and it also includes geography (placement on y-axis) and genomic relatedness (same colour = identical virus)
This reconstruction of a measles outbreak combines a contact network (lines) with an epidemic curve-style display (x-axis), and it also includes geography (placement on y-axis) and genomic relatedness (same colour = identical virus)

As I delved deeper and deeper into this emerging field of genomic epidemiology, two things happened that made me think it might be time to get back in the vis game more formally. First, as more and more genomic epi papers came out, I saw more and more terrible figures that violated basic principles of information display and design (function first, people, then form! Get thee to Tamara’s Vis Rules of Thumb right away!). With the exception of the absolutely stellar work of David Aanensen and Trevor Bedford (who, incidentally, are two of my FAVOURITE people in science), most genomic epi people were paying no attention to design and usability. As genomics moves closer and closer to the clinic, this is obviously Not Good At All. Second, I was fortunate enough to recruit Ana Crisan into my group, first as our bioinformatician/statistician and then as a PhD student (and Vanier Scholar! Go girl!). Ana had previous experience in using vis in the clinical space through her work at a local biotech diagnostics company, and her partner Matt Brehmer was, at the time, working on his PhD in Tamara’s lab (small world, eh?), so when it came time to establish her own PhD research trajectory, combining genomic epidemiology and infovis was the obvious choice. Tamara and I co-supervise Ana, whose work you’ll be seeing a lot of on this blog. In the meantime, follow her on Twitter for updates!

And that brings us to today. Ana is currently working on a very cool vis platform for exploring public health surveillance and lab data in an integrated, place-based fashion, and her first paper stemming from that work will be presented at BELIV 2016, a workshop held in conjunction with the IEEE Vis conference. Awesome Resident Geoff is working on the TB Report Design Project, which you can read about in the blog post he published yesterday. And I’m just kind of hanging out in the middle of it all, watching cool stuff unfold.

Genomic epidemiology has brought public health – a field that is usually pretty refractory to change – into a new era. With genomics and bioinformatics, we’re generating amazing datasets that are providing absolutely transformative insights into infectious disease epidemiology and control. BUT, all this awesome stuff is completely useless if we can’t communicate it to our end-users, the nurses, docs, and officials on the frontlines, many of whom may not know what a genome is or whose most complex computational task is sorting an Excel spreadsheet. That’s why I think one of the next logical steps for the genomic epi field is to stop making so many damned pipelines and start thinking about what comes out of those pipelines, visually. How do we report data in a meaningful and actionable way?

So, join our little group as we try to find a little bit of visual order in the data chaos that’s out there!

COMPASS-TB Report Design Study: Phase 1 Results

why-1352167_1280We have finished Phase 1 of our COMPASS-TB Report Design Study, an online survey exploring the utilization of different genome-derived data types for TB management tasks.

The full summary report can be found here.

Summary

Many public health agencies are starting to use whole genome sequencing as a tool for diagnosing infections, predicting drug resistance, and identifying closely related isolates that might suggest an outbreak. The COMPASS-TB study compared a novel WGS diagnostic pipeline to the standard diagnostic approach and demonstrated that, when this technique is used in the tuberculosis laboratory, we can generate all the usual results that one has come to expect from a reference mycobacteriology lab, but we can do so much faster and at lower cost (Pankhurst et al., 2016). As a result of this study, groups like Public Health England, the BC Centre for Disease Control, and the US Centers for Disease Control and Prevention are all using genomics to analyze their incoming mycobacterial isolates.

The COMPASS-TB study demonstrated that WGS can be used to diagnose, resistance type and genotype tuberculosis isolates more efficiently than the current diagnostic pipeline; however, further work was required to translate these often complex results into something that can be interpreted and used by clinicians and public health professionals. This study aims to use formal InfoVis methodologies to develop a report for the COMPASS-TB diagnostic WGS pipeline (Sedlmair et al., 2012). The design of the final report will be guided by user preference, while meeting both their needs and reporting requirements for accreditation (International Organization for Standardization, 2012).

In the first phase of the report design study, users of lab data were surveyed to assess knowledge of different data types and how they were used in routine clinical and public health management of tuberculosis.

While the number of participants was limited (n = 17), the survey  did provide some insight into what tasks genome-derived data may be considered useful. These insights will help guide the next phase of the study, where prototype reports will be developed in collaboration with human computer interaction researchers and evaluated through a second online survey.

Stay posted for our recruitment notice for the second survey!