Ensuring High-Quality Data for Rare Disease Research

At AllStripes, our mission is to unlock new treatments for people with rare diseases. One of the ways the AllStripes research team contributes to this mission is by asking important questions about the diagnosis, symptoms and progression of rare conditions. We do this by analyzing de-identified information extracted from participants’ medical records, looking for new patterns or connections. But as with so many endeavors, the quality of our output (research findings) can only be as good as the quality of our input (medical data). So before we can begin asking new questions about the rare diseases we study, we must first ensure our starting data is as complete and deep as possible. We recently detailed this process in a research poster presented at the 2020 International Conference on Pharmacoepidemiology (ICPE), using the rare condition chronic inflammatory demyelinating polyneuropathy (CIDP) as a case study.

What does it mean for research data to be “high quality”?

In the case of medical records, high-quality data would include as much of a participant’s medical record as possible, especially those documents related to care for the participant’s specific condition. Before we start research on a condition, we assess the completeness of each participant’s medical records using a variety of metrics, including the following:

• Record receipt: What percent of medical facilities listed by the participant have provided medical records?
• Recency: Does the participant’s medical record contain documents from within the last year (if the participant is living)?
• Gaps in care: Does the participant’s medical record contain any substantial gaps in time?
• Key clinical documents: Does the participant’s medical record contain documents critical to understanding the diagnosis, progression, or management of a specific condition? For example, does a participant with CIDP, a neurological condition, have neurology notes or neurological testing in their medical record?

Designing the study

For this case study in CIDP, we included participants who had consented to research and whose medical records contained documents from a majority of healthcare facilities listed by the participant. We also ensured that each participant’s medical record contained at least three neurology notes (because neurology is the most important specialty in CIDP care).

• 40 participants met the study inclusion criteria.
• The typical (median) participant had data from 59 clinical documents, including 13 neurology notes, from 4 different healthcare facilities.
• 10% of participants had records from at least ten facilities.
• 25% of participants had more than 100 clinical documents.
• 85% of participants had medical records from before they were diagnosed. These pre-diagnosis documents are especially important for helping us understand the early stages of a condition, even before it was recognized by a physician.

Evaluating CIDP symptoms and diagnosis

High-quality data from medical records should also contain information on key topics, including disease symptoms and diagnosis. To evaluate the depth of data available for this cohort, we confirmed the presence of important CIDP symptom information in participant medical records. We also investigated the time lag between when symptoms began and when patients were diagnosed with CIDP.

• All 40 participants had information related to CIDP symptoms in their medical records.
• Classic symptoms of CIDP include numbness, muscle weakness on both sides of the body, impaired balance or coordination, and pain. Each of these symptoms was experienced by at least 85% of participants (see below).
• The age of symptom onset was available for 33 participants. The typical (median) participant started experiencing symptoms at the age of 45. In contrast, the typical (median) participant in the cohort didn’t receive a diagnosis of CIDP until they were 49 years old, four years after symptom onset—underscoring the potentially lengthy nature of the diagnostic journey in this condition.

High-quality research in rare disease

Medical records represent a rich source of data for investigating the symptoms and progression of rare conditions and their impact on patients and families. But in order to conduct impactful research on these questions, researchers must start with complete medical records of appropriate depth. The AllStripes team has developed a process to ensure that the data we use in our studies meet these crucial criteria, empowering us to test important hypotheses about a variety of rare conditions. These quality assurance procedures will help us continue our mission of partnering with patients and families to accelerate therapeutic development in rare disease.