Iteration 3 Visualization Requirements

Iteration 3 - Exploring Reporting

Purpose / Preamble

This document captures a set of high level requirements for visualization / reporting from the Hub. We will use these requirements as we explore available modules and tools that could be used to enhance the Hub component. Enhancement is not in scope for this iteration, but the environmental scan is in scope, thus the need for some initial working requirements.

Types of Visualizations

There are many options for visualizations and ways to categorize the approaches. For SCOOP, it will be important to provide a core set of visualizations that are easy to use, as not everyone will be a researcher who uses the system. For more specific and complex visualizations, an export tool should be available. Instead, SCOOP will focus on doing a core set of visualizations that will support the majority of users an an intuitive manner.

For SCOOP we will consider the following visualizations:

• Trending over time of single and multiple variables
• Scatter plot with trend lines of two items (e.g. # of diabetics in a practice and average A1C per practice)
• Scatter plots with 3rd variable represented by size / colour / shape of indicator.
• Comparison of different categories -- bar charts
• Visualization of data quality in an answer
• Venn Diagrams?

Visualization Features

Feature: xx-xx TITLE In order to visualize relative quantities The question manager should be able to Plot multiple results of a question set as a bar graph.

Feature: xx-xx TITLE In order to visualize temporal trends in data The question manager should be able to plot results of a question over time, either from a single query set or from repetitions of a query across the network.

Export Features

Feature: xx-xx TITLE In order to share findings in a report or publication The question manager should be able to Export a visualization as a PDF

Out of Scope Visualization

For the current focus, we will assume that qualitative data visualizations are out of scope. Flowcharts, gantt charts, and other workflow / dependency charts are not in scope. Also, animated trends over time, while they can be powerful, are not needed at this time for SCOOP.

Data fundamentals for constructing visualizations

The following topics will be addressed in the sections that follow:

• Common relationships in quantitative data
• Visual options for encoding quantitative data
• Best and worst practices for representing select data relationships

Following the discussion of visualization strategies and tactics, an overview of open source visualization technologies will be provided, complete with a sampling of representative visualizations.

Common relationships in quantitative data

Quantitative data consists of numbers. When visualized, quantitative data is usually depicted relative to some other category of interest. These categorical descriptions, or scales, come in three flavors:

• Nominal - Categories of data that aren't really related, but instead just described by name (e.g. Doctor A, Doctor B, Doctor C)
• Ordinal - Categories of data that are related in that they follow an implied order, although they aren't really quantified (e.g. underweight, normal weight, overweight)
• Interval - Categories of data are related, ordered and represent quantitative values themselves (e.g. BMI < 15, BMI >= 15 and <25, BMI >=25)

The following list describes 7 common relationships that graphs commonly depict. When selecting a graphical design, it is best to start by identifying the relationship you'd like to communicate between your quantitative data set and the select categories of interest.

1. Nominal Comparison

Here, you just want to place categorical data measures side by side to facilitate comparison, with no desire to encourage ranking or identification of larger data patterns.

• Number of patients per doctor
• Number of tests per patient

2. Ranking

Here, you want to facilitate more than just a comparison of nominal categories. You want to draw attention to the relationship that exists between a category and its quantitative value in some order (e.g., from smallest to largest).

• KEYWORDS: larger than, smaller than, equal to, greater than
• Number of patients per doctor, ordered by doctor with most to least patients
• Number of telephone consultations by clinic, ordered by clinic with most to least consultations

3. Time Series

Here, you want to depict the relationship of quantitative measures as they were captured in time.

• KEYWORDS: change, rise, fluctuate, decline, grow, trend
• Billing revenues per month
• Average number of minutes between check-in and examination per hour of operation
• Lab values at given times of day

4. Parts-to-Whole

Here, a set of nominal, ordinal or interval categories are depicted so that it becomes possible to see what proportion of the total quantity is represented by a given category.

• KEYWORDS: rate of total, percentage of total, share of total, accounts for X of total
• Percentage of clinic revenue earned by clinic physicians
• Percentage of clinic patients according to age bracket

5. Deviation

Here, you are interested in illustrating how a set of quantitative values belonging to certain categories differ from a set of reference values.

• KEYWORDS: variance, difference, plus or minus, relative to
• Average difference between target weight and actual weight in patients from three different intervention groups
• Degree to which systolic blood pressure measurement changed compared to previous month's value

6. Distribution

Here, you want to depict how quantitative values were observed across an entire range of ordinal or interval categories.

• KEYWORDS: frequency, distribution, range, concentration, normal curve
• Number of patients with a chronic condition, in ascending age categories
• Number of medications per patient, in ascending intervals of BMI

7. Correlation

Here, you want to illustrate how two sets of quantitative data relate to one another, perhaps to look for a possible causal relationship.

• KEYWORDS: increases with, varies with, follows, affected by, caused by
• Correlation between serum hemoglobin and ferritin lab results in 1000 patients

Tips for encoding select quantitative data relationships

Before talking about specific graph types, here is a quick list of the visual objects commonly used to encode quantitative values:

• Points
• Lines
• Vertical Bars
• Horizontal Bars
• Vertical Boxes
• Horizontal Boxes

When designing a graph to represent a given data relationship,