Adam Sage

Presentation: http://www.slideshare.net/secret/FV6UWL51XGRqDu

PROPOSAL:

For my midterm project, I want to develop several options for visualizing adherence to antihypertensive medication and the physiological effect as measured through blood pressure. For this, I will take data from a week of daily adherence to a medication, dichotomized as yes/no for each day, as well as diastolic blood pressure, systolic blood pressure, mean arterial pressure (MAP), and an algorithm I have developed which combines MAP and adherence into an individually tailored single composite measure and is recalculated for every new reading of daily adherence and blood pressure.

The goal for this project is to develop several versions of visualizing different combinations of this information, which I could then present to patients in focus group sessions and cognitive interviews to assess preferences and comprehension. Ideally, I would be able to then use this information to develop experimental conditions which would allow me to tease out what information, when presented as a visualization, facilitates comprehension of how antihypertensive medication affects blood pressure. This is an important research question as knowledge is known to directly affect a patient's self-efficacy, which is an important factor in medication adherence.

so perhaps two sets of interfaces:

  • consumer facing (patients)

  • experts (understand how antihypertensive medication affects blood pressure)

PROJECT:

The goal of this project is to develop a series of visualizations to communicate medication adherence and the physiological effect of blood pressure for patients with hypertension (high blood pressure). Presumably patients not taking they medication appropriately would experience elevated blood pressure readings. Thus, it’s important to develop visualization techniques for providing patient feedback that allows patients to comprehend how their medication adherence effects their blood pressure. However, this is a task that requires patients to interpret three pieces of information: 1) medication adherence, 2) blood pressure, and 3) the assumed link between taking medication and controlling blood pressure.

I have created a dataset with hypothetical patient data on medication adherence and blood pressure for seven days, which I used to develop a series of visualizations which I hope to eventually test with real patients in focus groups, cognitive interviews, and eventually an experiment that allows me to formally test and determine which visualization techniques result in the best comprehension of health status and the benefit of appropriately taking medication.

Context:

The context that users would view the visualizations for blood pressure and medication adherence would most likely be mobile apps, and possibly patient portals (e.g., Health Heels Portal).

User:

The intended user would be a patient, so the visualizations must be easier to understand for a broad spectrum of individuals with regards to characteristics such as color blindness, cognitive processing limitations, hypertension knowledge, health literacy, eHealth literacy, numeracy, and graph literacy.

Task:

The task for these visualizations is to comprehend three specific pieces of information related to antihypertension medication and its physiological effect:

  1. Medication Adherence – whether a patient took they medication appropriately (dichotomized as yes/no for each day)

  2. Blood Pressure – systolic, diastolic, mean arterial pressure (via MES)

  3. Linking Medication Adherence to Blood Pressure – communicating appropriate medication adherence results in improved blood pressure control

Dataset:

The data source for these particular visualizations is hypothetical medication adherence and blood pressure readings across seven days. I have medication adherence dichotomized yes/no, systolic and diastolic blood pressure, mean arterial pressure calculated using the blood pressure measures (this is a single value for blood pressure rather than two values), and a calculation of medication effectiveness score (MES), which is the algorithm I created for tailored feedback (this is also a single value for blood pressure, but it is tailored to reflect how medication uniquely effects a patient).

Visualization Technique:

Because I am interested in communicating feedback over time, I needed to show a trend, which is best illustrated using line graph. I plotted the interval variable of “Day” across the X-axis, and varied which continuous variable to plot along the Y-axis. These continuous variables included 1) systolic and diastolic blood pressure, 2) MES, and 3) MAP. In addition, I needed to illustrate medication adherence, which is a dichotomized categorical variable, as well as systolic and diastolic blood pressure for graphs using MES and MAP along the Y-axis. To illustrate medication adherence, I chose two techniques 1) color-coded line segments with blue indicating a missed dose, and orange indicating adherence, and 2) symbols indicating whether medication was taken for that day. To illustrate systolic and diastolic blood pressure, I chose to include those measures below each day because these are secondary pieces of information that supplement the trend line. Labeling points for each graph was also varied.

I primarily chose gray scale colors when color was not needed. Where I used colors, I chose not to use very bright and bold colors, and consider users that may be color-blind.

I used Tableau to create the initial graphs, then made small edits using an image editor.

Display:

With respect to display, these images are not formatted to fit a smartphone, but they are perfect for a computer display. Adjustments would need to be made before implementing any of these visualizations in a mobile app.

What characteristics must the data visualization have?

To begin, I considered my data types and prioritized them. I also considered what techniques to visualizing these data types are acceptable (not necessarily better than the others). For instance, I did not use bar charts because they are not the best method for illustrating trends, and blood pressure is a single measure, not a proportion. Based on the data types, I determined that at minimum, the visualizations must be line graphs with the interval variable “Day” along the X-axis, and continuous variable for blood pressure along the Y-axis.

What aspects of the visualization technique could be altered

I next looked at what aspects of the visualization I can vary. I decided the following aspects should be presented to patients in order to formally test preferences:

  • Medication adherence (symbols vs. color-coded line)

  • Labeling points

  • Blood Pressure measurement type

Do you agree or disagree with any of the decisions I made?

Some characteristics I made and “executive decision” in terms of technique I used, but do you think any of these or other are worth testing?

  • Color choices

  • Size of the lines

  • Width and Height of the graphs

Is there something missing that I did not consider?