Melissa the Clinical Analyst

  • Melissa needs to access data about clinical operations and quality-related incidents to assess and improve clinical care.
  • She struggles to access additional datasets needed for reporting and to make sure all stakeholders are aligned on data definitions.
  • We can help her by facilitating access to raw or external data and by providing her team a better way to create and document curated datasets to monitor operations and quality indicators.

Melissa needs a unified data source and code repository to make sure everyone is using the right “source of truth”

Melissa wants to help things run smoothly at Fred Hutch, and she does that by monitoring clinical operations and reporting on quality improvement strategies. What percentage of patients have filled out patient-reported outcomes (PROs)? How many patients are referred to Fred Hutch for cancer treatment? How many medical errors or near misses have there been in the last month? Melissa needs to answer all these questions and more, pulling data from multiple systems like the EHR, lab data systems, cancer registry data, and even data from phone systems. With so many data sources structured in different ways, there are many possible “sources of truth”. Melissa needs to both define what she’s looking for and figure out where to get it. But getting data from so many different sources can take a lot of time.  Melissa needs data in a central platform, structured in a way that makes sense for her work. Since she’ll never get all the data she needs in one place, she needs to be able to easily join curated datasets to raw data from Epic’s warehouse, or even to excel sheets of chart abstracted data from stakeholders, as needed. And after all the work that goes into defining what data to get, Melissa needs a central repository for code snippets and the definitions and documentation associated with data objects so she can share them with her colleagues and make sure everyone is using the right “source of truth” for a given project. It would also save her time if she had more tools to automate parts of her workflow.

Collaborators: Other roles on the clinical analytics team such as epic report analysts, Alex the BI/Analytics Engineer, Daisy the data scientist

Downstream users: Super Users, Bashir the Business Analyst, Clinical Leaders

Key Challenges

  • Identifying the most accurate source of truth when multiple options are in play
  • Needing to import data from other systems to do her work
  • Understanding what all data is available to import when it is in systems that she does not have access to
  • Ensuring that all stakeholders both within and outside of the team are using the same reference data and data definitions, and resolving confusion when discrepancies occur
  • Identifying what epic fields to query when there is more than one option (e.g., more than one comment field for clinical comments), or when official data dictionaries differ from how data is entered in practice
  • Providing stakeholders with what they need with a very short turnaround time

Needs and Wants

  • PHI-approved space to easily access different types of data, perform cloud-based analysis, and enable the building of dynamic visualizations
  • Automated reports on frequently needed data
  • Availability of clear, user-friendly documentation within the data platform, including information about known data issues
  • Repository for code and data object definitions, including information about when the use of an EHR field differs at Melissa’s institution from its definition in the standard dictionary
  • Ability to share curated datasets and appropriate documentation with selected stakeholders from a centralized platform to enable some limited self-serve data access for external stakeholders 
  • A way to clearly communicate data definitions to end users

Data types used

  • EHR data
  • Cancer registry data
  • Lab data and radiation oncology data
  • Clinical trials management system data
  • Medication and other treatment data
  • Data from internal systems (e.g., phone systems)
  • Externally sourced data (e.g., patient satisfaction surveys, market data, ICD and CMS codes, public datasets)
  • Custom reference data (e.g., institution-specific classifications)

Image attribution: “Woman In Black Blazer Sitting On Black Office Chair” by quariesofficial is licensed under CC BY 2.0.

Acknowledgments: Thanks to members of the clinical and financial analytics and data platform teams for their input!

last updated July 2024