New Atlas Features for UCB: Cohort Sample and Profile Timeline

Developing patient-specific capabilities under an open source license

On behalf of global biopharmaceutical company UCB, The Hyve has developed new capabilities in Atlas, the analytics component of the OHDSI suite. We developed a new cohort sampling function and an extended patient profile view. In line with our open-source policy, we shared the newly developed features with all members of the OHDSI community via the main Atlas and WebAPI repository.

The challenge

Atlas is a powerful tool for defining observational studies on OMOP CDM data. Defining appropriate cohorts of patients with characteristics of interest is a key step in doing observational research. Our client wanted to extend the capabilities of Atlas to enable exploration of the defined cohorts on a patient level. In this way it becomes easier to verify whether a cohort definition produces the desired result.

How we solved it

New cohort sampling function

To meet our client’s needs, The Hyve’s developers started by adding a Samples tab to the Cohort page in Atlas. This allows users to create samples of any size for a selected data source, with the option to refine a search based on gender and age criteria. Clicking on the generated sample will show the list of patient IDs included in the sample and characteristics like gender, age and number of events. Each individual in the sample can then be further explored using the patient profile link. Furthermore, generated subsets persist until a new cohort has been created for the selected data source. One benefit of this feature is that it allows for collaborative analysis of the cohort sample investigation and reinvestigation of the subset later in time.

Extended patient profile view

Our developers reworked the patient Profile tab to enable a user-friendly exploration of a patient’s medical history in a more targeted way. Some of the new features we added to the existing profile:

• Each unique concept is now displayed on a separate line. For instance, all exposures to a particular drug are on one line. The lines can be collapsed as well, to e.g. show all drug exposures of a person. This is implemented for each of the OMOP clinical domains (condition, procedure, observation, drug, device, measurement). In the screenshot below the occurrences of ‘Malignant tumor…’ are shown.
• Concepts (a particular drug, condition, observation, etc.) can be pinned/unpinned to create a focused view of the patient timeline. Therefore, investigators can create a custom view of the most relevant events, easing interpretation of the events.
• For the purpose of cohort exploration, we have developed the option to display up to 10 patients in a similar view (stacked one below the other). This allows investigators to easily compare multiple patient timelines in a data source/cohort. This feature, combined with the concept pinning one, allows for a deep exploration and comparison of patients within a cohort.
• The time unit can be toggled between relative and absolute dates. This feature is particularly useful when there is some external event influencing the data. For instance a drug being withdrawn from the market or disease coding changed.