`

Results

The Challenges of Disparate Data Formats: Analysis and Visualization in the SLIDES Project Ang Li1 ● Steven Chall, MS2 ● Sherry Wenshun Liu1 ● Allison Vorderstrasse, DNSc1 ● Constance M. Johnson, PhD1

1 Duke University, Durham, North Carolina 2 Renaissance Computing Institute, Chapel Hill, North Carolina

To evaluate preliminary effects of VE participation on metabolic outcomes in the SLIDES

study, we developed an Internet visualization tool with a user-friendly interface for

dynamic processing of multidimensional data.

This visualization tool was designed to accommodate data in disparate formats.

Multidimensional Process Data from SLIDES Project

We collected multidimensional data over a period of 6 months.

Quantitative data included but were not limited to:

• Movement, interactions with objects and other participants, proxemics

• Time spent in the site, frequency of log-ins

Qualitative data included: Observational data, voice, text, email, forum, focus groups

Visual data included: Photos, videos

Quantitative variables included:

Demographics

Perceived usefulness

Perceived ease of use

Presence

Co-presence

Metabolic indicators: BMI, HbA1c, BP

Diabetes knowledge

Self-efficacy

Self-management behaviors

• Diet, exercise, foot care, glucose testing, smoking

Perceived support for diabetes management

• Professional, family, friends

Initially, visualizations were created using the

R environment (open-source data analysis

and statistical computing software).

Visualization capabilities were extended by:

Incorporating R-Shiny libraries (which include

a larger repertoire of visualization approaches

plus web deployment).

Applying D3.js JavaScript libraries (which add

capability to animate results, as well as more

versatility and power).

User interface includes an animated dashboard

(shown at right).

Provides multiple views of multidimensional data beyond the traditional 2-D perspective.

Using color, shapes, volume, and animation facilitates exploration of the data through visual images.

Dynamic drawings can uncover evolutionary paths of change over time.

Animation allows combination of multiple types of data, highlighting temporal aspects of the data.

Visualization: Why?

Text transcribed from

audio recordings of

participant conversations

Survey

results

Biometric

measurements

Numerical

avatar positions

within the VE

Images

video

&

Visualization

Tool

Sees New

Thematic

Patterns

in Data

The dashboard which shows different visualizations

of our data are animated.

Moving the mouse over the various visualizations

provides different views of the data on a single

screen, yielding more in-depth understanding.

If a user wants to view just one or two visualizations,

the functionality of this dashboard allows the user

to drill down to look at the specifics of the data.

The Dashboard Visualization: How?

Methods: Analysis and Visualization

Development of Visualization Tool

`````

Visualization allowed us to see through animation how each subject’s weight, HbA1c

(indicator of metabolic control), and corresponding VE activity level changed over the

course of the study (6 months).`

Results

Heat Maps: Visualization of Locations in VE Visited by Subjects

Animated Visualization of Changes in

Subjects’ Weight, Hb1AC, and Activity in Virtual Environment

Animated Visualization Showing the Amount of Participation

by Each Subject at Each Location in the Virtual Environment

This visualization shows that Subject 7

clicked on objects in the site 386 times

during the 6-month SLIDES study.

S13:

90 times

S7:

386 times

Subjects Subjects

Objects in

the virtual

environment

Objects in

the virtual

environment

This visualization shows that Subject 13

clicked on 90 different food items over

the course of the 6-month SLIDES study.

``

The sequence of screen shots above is from an animated browser-based visualization

which shows these characteristics for SLIDES at baseline, 3 months, and 6 months.

In each graph, the horizontal axis is weight and the vertical axis is HbA1c.

Each subject is represented by a circle with a unique color, so individuals can be followed

over time.

When the mouse is clicked over any circle, the visualization will identify the subject.

The area of the circle represents the amount of that person’s participation in the site.

Screenshots below show that Subject A decreased her HbA1c and lost weight over the 6 months.

Weight

Hb

A1

c

Weight Weight

Hb

A1

c

Hb

A1

c

Baseline 3 months 6 months

Heat maps show the locations

within the virtual environment

that were visited by SLIDES

participants within a single week.

The heat map is not automatically

animated; use slide bar at bottom

to select a week for analysis.

Screenshots of heat maps that

show locations visited by

participants during:

• first week of the study (“0”, left)

• first week of the second half of the study (“12”, right).

`

`

Type 2 Diabetes (T2D) is a chronic disease epidemic in the U.S.

Individuals with T2D provide 99% of their own care.

Self-management includes diet, exercise, glucose testing, etc.

Self-management goal for persons with T2D: Achieving metabolic control, which reduces the

morbidity and mortality associated with this disease.

Achieving this goal through self-management is challenging for many persons with T2D.

Innovative interventions to empower patients in diabetes self-management are

needed.

In response to this need, we developed and tested the SLIDES Project, an intervention for

adults with T2D which uses a virtual environment (VE) to provide education about diabetes

self-management and ongoing support.

To analyze the multidimensional data collected during a 6-month test of this intervention,

we employed innovative visualization techniques.

SLIDES (eHealth: Second Life Impacts Diabetes Education and Self-Management)

This intervention uses a virtual environment (VE) to provide ongoing education and

support to adults with T2D.

SLIDES Program was designed within the popular online multi-user platform Second Life.

The VE was available 24/7 on the Internet, minimizing access barriers.

The SLIDES Program offered:

Weekly synchronous diabetes education classes

Group meetings

Social networking in a virtual community in which participants practiced real-world skills such as

grocery shopping, exercising, and dining out, allowing for interactive knowledge application.

SLIDES Study Design

One group, Pre-Mid-Post Study

Phase 1: First month

Oriented to the site

Baseline survey administered

Metabolic information (HbA1c, height, weight, BP)

Visit site 2x / week (classes, interactive sessions, other)

Phase 2: Next five months

Visit site at will (classes, interactive sessions, other)

Survey and metabolic information at 3rd month

Phase 3: Sixth month

Survey, metabolic information, and focus groups

The SLIDES community

Background

` SLIDES Project: Intervention for Adults with T2D

SLIDES Study Sample

20 participants with Type 2 Diabetes

21 ‒ 75 years old

Computer and internet literate

Broadband connection

No severe diabetes-related complications

or late-stage chronic disease

Graphics can be combined

to compare changes over time

in variables with disparate

data formats.

Composite graphics at left

allow comparison of objects

touched to changes in

subjects’ visits to VE sites.

Combining Graphics

BMI Change Over 180 days

Conclusions

Changes in Self-Efficacy and

Social Support Over 6 Months

Hovering the mouse over each line shows the

participant and change in BMI over 6 months

This animated browser-based visualization

shows how self-efficacy and social support

changed from baseline to six months. These

changes were statistically significant.

Further Examples of Visualization

Self-Efficacy

So

cia

l S

up

po

rt

Baseline

6 months

We developed an internet data visualization tool for analysis of data in disparate formats

from the SLIDES Project.

The tool generates a variety of visual representations to facilitate recognition of patterns

and temporal relationships in the data.

These visualizations help to show how data clusters, which in turn helps researchers to

see new dimensions in large and diverse data sets

Spatial transformation of data collected from SLIDES participants in disparate formats

revealed trends and phenomena that could not be identified using traditional graphs.

Visual representation of the data with animation made salient information more apparent,

and enhanced our understanding of how time and group dynamics may affect

self-management of T2D.

Major challenges remain in integrating the wide variety of data gathered in the SLIDES

study, including problems with rendering raw audio data into a format suitable for analysis,

extracting meaning from the resulting text, and especially performance.

Finding solutions to these problems could be beneficial to researchers who use mixed

methods approaches to analyze healthcare “big data”.

As the interactive web becomes a popular mode of delivery of health information, we need

interactive research tools to better understand the outcomes of these studies.