Mapping Clinical Data To Well-Being Domains: Epilepsy Wellness Wheel

Summary

A broader model of health that includes patient functioning across life domains is becoming increasingly common in disease management. The Epilepsy Foundation’s Wellness Wheel is one such model containing a wide range of domains. To help healthcare providers better grasp this expanded picture of the individual, there is a need to map and report clinical data organized around life domains. This article describes the implementation and methods used to (1) map clinical data into appropriate life domains, (2) calculate common metrics across domains, and (3) report results that can be quickly and accurately understood.

 

Aiming Across Domains

One of the most consistent findings among individuals with chronic conditions is that enhanced functioning across life domains plays a critical role in the control and treatment of disease. This has prompted numerous initiatives designed to promote and advance a well-rounded approach to disease management. One example is the Epilepsy Foundation’s Wellness Wheel model which organizes contributing factors related to overall health and well-being into eight domains (e.g., sleep, stress, emotion, diet, etc.). Epilepsy clinics are moving in sync by broadening data collection efforts across Wellness Wheel domains.


There is parallel movement away from the traditional way to manage these data using a siloed database with analysis in secondary systems, to Learning Health Systems that dynamically leverage data real-time to assist providers with clinical decision making, and engage patients and their caregivers by providing personalized information, incentives, and a secure provider communication platform (e.g., “Medication Adherence: Toppling Barriers With A Learning Health System”). This article describes the implementation and methods used to help healthcare providers better grasp patient functioning across a broader range of domains. The aims were to address three needs:

  1. More formally organize clinical data around Health and Wellness domains.

  2. Develop common metrics for within and across domain comparisons.

  3. Ensure report results are organized around domains and can be quickly and accurately understood.

Puzzle Pieces

When first dumped out, it is hard to appreciate the pile of puzzle pieces form the picture on the box when arranged properly. So too each clinic visit produces a pile of data in need of proper arrangement to see the picture of patient functioning across life domains. The range of domains have been previously discussed (“Disease Information: Making It Relevant To Patients”) and are listed here in name only:

  • Seizure (e.g., type, biological markers, etc.)

  • Treatment

  • Social Determinants of Health

  • Self-Management

  • Emotion

  • Stress

  • Sleep

  • Diet

  • Independent Living

  • Social

  • Physical Activity

  • Education and Work

  • Cognitive Function

Although the domains may be neatly organized conceptually, the process of translating raw data into a clinical picture of functioning is complicated by many factors. For instance, there are often numerous sources of clinical data including the patient, family members and/or caregivers, various provider specialties (i.e., neurology, neurosurgery, social worker, nurse, occupational therapy, etc.), other systems (EMR), devices, etc. Indeed, epilepsy assessment often requires multiple sources as patients may have limited awareness of seizure characteristics. Also, data from each source may have a different levels of measurement (nominal, ordinal, interval and ratio-level data, see here). In other words, the data often do not share a common metric or scale (e.g., ‘Moderate’ = 5), complicating comparisons and ruling out simply summing things together. In addition, data may be applicable to more than one domain. Consider a depression questionnaire that asks about sleep, emotional state, diet, physical activity, cognitive function, etc.


It is important to also note that even when these complicating factors are addressed, it is easy for providers to get bogged down in a sea of numbers when attempting to grasp the full picture of results across domains. In other words, how do you keep results from looking like a pile of puzzle pieces? To overcome these complicating factors, there are three main tasks involved.

  • Map: Determine the domain to which each data element will be mapped

  • Score: Generate common metrics that apply to all data elements

  • Report: Create reports that allow results to be quickly and accurately consumed

Connecting Pieces: Map

Identifying the appropriate domain for each item is the first step. An expert consensus approach was used, and agreement was found for over 95% of item / domain pairs. Disagreements were resolved by a senior-level neurologist. Items from all sources (e.g., patient, caregiver, provider) were categorized.


There was a good bit of variability in the number of items and the scale of measurement within each domain, highlighting the need for a common metric within and across domains. For example, the Emotion domain had over twice as many items as the Sleep domain. As a side note before transitioning to developing metrics, seeing the item distribution across domains made it easy to identify areas of assessment strengths and weaknesses. The mix of strengths and weakness is neither positive nor negative, just depends on the aims. However, it can be helpful information when considering changes to the data definition (e.g., adding or removing items).


Connecting Pieces: Score

Two scoring methods were created, both of which started by assigning a negative to positive valence to each item’s response set (see image). Then, the values zero and two were assigned to the positive and negative anchors, respectively (i.e., higher values associated with poor functioning). All other responses were assigned a value of 1.

There were two items exceptions to the scoring method (about 10 items total).

  1. Items without a clear negative/positive valence were excluded (e.g., Abstinence method for birth control).

  2. The positive valence anchoring point was adjusted to include more than one response (e.g., Last Seizure: ‘Today’ [negative anchor], ‘1 to 6 days ago’, ‘etc.’…, ‘1 to 2 years ago’, ‘More than 2 years ago’ [the last two were both considered positive anchors]).

With the item scores in hand, two metrics for each domain were calculated (see image).


Connecting Pieces: Report

Report design principles have been previously discuss, see “Patient Registries: Clinical Report Design Techniques.” The main techniques to be described here include: General-To-Specific Layout, juxtaposing sources, images, and color. Briefly, the report was structured beginning with high level summary information at the top (General) followed by increasing detail moving down (Specific). This allowed the consumer to quickly identify any problems (or strengths) and move into details as needed. Also, the source data for any given domain was juxtaposed to make it easy to appreciate the full picture of functioning and spot discrepancies. Graphic images were used to make it easy to find domains of interest. Dynamic images that combined color and a visual score representation made it simple to grasp results. As an example, consider the use of a series of speedometers going from green to red, each with an associated dial location (see image). Only the appropriate speedometer appears.

All these concepts are on display in the domain map below. Notice Patient and Provider responses are juxtaposed, the overall general summary (i.e., speedometer) is at the top, and each domain has an image with the background color corresponding to the percent maximum score. Note, Social Determinants of Health (SDOH) and Self-Management domains are only related to the patient.

The next section in the report contains the details for each domain, again with the Patient / Provider juxtaposition.

  • The domain image is used for easy identification.

  • A proportional horizontal bar chart displays the percent endorsement within range score value (i.e., 0, 1, 2) and is color coded. A frequency item count is also displayed in the bar chart as additional information for interpreting results.

  • A dynamic color-coded dial indicates the percent maximum score.

  • Individual items and responses are listed in a color-coded table.

  • Positive anchors (green) items are listed as a long string, separated by a semicolon.

  • Negative anchors (red) and items with the value 1 (yellow) are listed in bullets to facilitate case conference discussion.

Finally, some domains lend themselves to a visual detail approach, rather than text-based. For instance, consider two items from the Social Determinants of Health domain (shortened for brevity).

  1. What is your housing situation? __Stable __I’m worried about housing __No Housing

  2. Are you safe where you live? __No __Yes

Combining color coding and icons, all six possible permutations of the above two items can be created and dynamically displayed as appropriate (see below). Thus, even in the “Details” section text can be eliminated and information quickly consumed!

Taken together, these methods provide a foundation for mapping clinical data to appropriate domains, and a quick and easy way to interpret results. Note that the domain mapping described is just one section to a provider targeted report incorporating many other elements to assist with clinical decision making. Similar principles are being applied to deliver personalized information to patients (learn how here). Lastly, these data are housed within an integrated research application (e.g., clinical trials, observational studies, surveys, registries, etc.) to facilitate across project collaboration. Much more to come…

Recent Posts

See All