Tremor Registry: Shaking Up The Data Work Camp
This article describes steps taken to transform a patient registry exclusively used for disease natural history academic publications into a clinical tool that facilitates day-to-day operations and engages patients in the treatment process. This case example used a long-standing (2+ decades) registry of patients with tremor disorder. This article examines issues surrounding the creation of single / multi-subject reports. Other topics will be covered in forth-coming articles.
Building A Wall
In Alexander Solzhenitsyn’s classic story, “One day in the life of Ivan Denisovich,” survival in the gulag work camps involved maneuvering to avoid the most treacherous jobs. Ivan manages to land one of the more desirable jobs, constructing a wall for a building. Just as walls separate inside space from outside elements, so too does patient registry design involve walling off what’s included and excluded across domains (see image). The aims of the registry are the primary sorting mechanism to determine inclusion status. But what if the aims change over time?
This is the first in a planned series of blogs describing a DIY ‘renovation’ centered around enhancing and re-shaping the walls of a tremor disorder patient registry. The main goals of the project were two-fold:
1. Improve day-to-day clinical operations and decision making
In the work camp Ivan risked severe punishment, even death, by hiding the highest quality tools for his own use on the job. The link between quality tools and quality work isn’t always direct, but the two are inextricably tied. The essence of a registry to function as an effective tool is the content. For this renovation project, the starting point included content across several domains with data captured exclusively by clinician-based ratings over a 20+ year period.
e.g., Onset date, initial tremor location
e.g., Other movement disorders
Most prominent and location (e.g., upper extremity)
Location (e.g., trunk, chin, voice, left / right, upper / lower extremity)
Type (e.g., action, resting)
Function (e.g., ability to draw / pour / handwrite / etc.)
Treatment and Procedures
e.g., Medication use and side-effects, surgical procedures
Activity of Daily Living
e.g., capacity for employment / independent living
Historically, these data were intermittently queried for analysis and publication (see example here), but were not used in day-to-day clinical operations, decision making or patient engagement / education. The renovation steps were designed to address this limitation and were completed in the following order:
Provide clinical utility of data “as-is” and integrate into the practice
Augment and/or add domains to the registry
Patient Reported Outcomes (PRO)
Imaging and electrophysiology (e.g., deep brain stimulation - lead location and stimulation parameters)
Bench data (e.g., genetics, proteomics, etc.)
Integrate registry changes into day-to-day operations
Better engage and inform patients in their care
Improve provider case management and decision making
Repeat steps for each movement disorder in the registry (e.g., Parkinson’s Disease)
Item 1 is the focus of this article.
Despite the work camp’s collectivist group-think and forced behavior, Ivan's work on the wall reveals unique craftsmanship and beauty, demonstrating the sovereign individual is not merely a handmaiden of the state. This creative spirit also guides efforts to transform a ‘bunch of registry numbers’ into something clinically meaningful and actionable. Clinical reports are a common way to do this. More detailed report design techniques are covered in other articles (see “Patient Registries: Clinical Report Design Techniques,” here). Instead, this article presents specific examples of what was created for the tremor disorder registry.
Single Patient-Level Reports
Graphical display can be a quick and easy way to grasp outcome measure scores over time (see image below). To ease interpretation, graphs were grouped by body location (e.g., left vs. right) and conceptual domain (e.g., drawing tasks).
Current Status, What’s Changed?
Graphs can quickly become cluttered and have limited ways to represent location information beyond just grouping locations (e.g., left vs. right). Moreover, multiple factors (e.g., disease stage, time since last visit) often make 'change since last visit' the primary clinical concern which requires close inspection and mental effort to determine from a graph.
In contrast, consider the visual images below. Admittedly, what comes to mind is a poor attempt at Minecraft characters.😊 However, embedding severity color coding into an anthropomorphic image makes change since the last visit immediately apparent. The image also seems less cluttered despite more assessment domains and information being displayed.
Other Ratings and Key Concerns
Assessment measures that did not fit the anthropomorphic image well (e.g., not tied to a location, different scaling, etc.) were displayed in a more traditional table format. Red versus green images and text color were used as negative (bad) / positive (good) indicators.
As an example of a general principle, note that a “Yes / No” response set has fewer display options than a variable ranging from 0 to 10. Again, the aims of a registry determine the content. However, measure selection balances a number of different factors (e.g., psychometric properties, assessment domain, etc.) that will be discussed in a separate article covering the addition of Patient Reported Outcomes to the tremor registry. For now, just note the data definition impacts display options. For instance, medication side-effects was a “Yes/No” question in the tremor registry, but note how adding two fields, 1) an indication list and 2) “Yes/No” rescue medication, can dramatically change the display options (see below; from an epilepsy registry).
Reports that summarize information across patients are also key to integrating registry data into actionable information within a clinical setting. Two examples are provided below, other examples will be forthcoming in a detailed publication.
Annualized Change Scores
An annualized tremor total change score was generated for every clinic visit and compared to the initial (i.e., baseline) visit. The change score was then compared across various patient characteristics. As an example, age group differences are shown below.
For each medication in the tremor registry, the percentage of visits for which a medication was discontinued due to side-effects was documented. As with total tremor change score, medication discontinuation can be separated out and examined by a number patient and disease characteristics. As an example, an aggregate summary of the top five medications prescribed for tremor are presented below.
Almost A Happy Day
At the end of the long day described in Solzhenitsyn’s book, Ivan wonders whether it was ‘almost a happy day.’ Given the shocking and brutal events of the day, it’s a damning indictment of the misery and psychological torment of life in a work camp. Although renovations often involve unhappy days, there are many incremental improvements leading to happier days. There is ongoing work to the tremor registry that will be described in subsequent articles, and eventually there will be a ‘happy day’ when everything is pulled together! More to come.