Content
Summary
Discrete laboratory data is structured information that enables efficient clinical decision-making and research, contrasting with unstructured formats like PDFs that hinder automation. Implementing this requires standardizing data through synoptic reporting, establishing interfaces between Laboratory Information Systems and external labs, and utilizing Natural Language Processing to extract key concepts. The FLT3 mutation case study illustrates how moving from generic attachments to specific, searchable fields improves both clinical workflow and data utility.
Intuition
Think of laboratory data as a library. If books are thrown randomly on the floor (unstructured PDFs), finding a specific story takes forever and prevents you from analyzing trends across all books. Discrete data is like a cataloged library where every book has a specific shelf and tag. This structure allows physicians to instantly see if a patient's condition is improving or worsening by comparing specific data points over time, rather than manually reading every document. To achieve this, hospitals must not only organize the data but also ensure that external labs send information in this organized format, often using technology to read and categorize unstructured text automatically.
Derivation
- Identify the problem: Legacy systems store lab results as non-discrete data (PDFs, free text), causing inefficiencies and barriers to data integration.
- Define the solution: Transform data into discrete formats using informatics principles, NLP, and machine learning to extract medical concepts.
- Implement structural changes: Adopt synoptic reporting to standardize result formats and create interfaces between LIS and external reference labs.
- Engage stakeholders: Involve clinical, IT, and laboratory staff to map workflows and ensure the new architecture meets both clinical and administrative needs.
- Validate impact: Ensure clinicians can view results efficiently, trend markers over time, and that data is available for translational research and precision medicine.
Worked example
At an academic cancer center, clinicians struggled to access FLT3 mutation results for hematology patients because results were stored as scanned PDF attachments under a generic 'miscellaneous' category, preventing automated searching and trending.
- A team reviewed the workflow and identified the need for discrete data fields.
- A discrete orderable test for 'FLT3 mutation' was created in the CPOE system.
- Results were configured to file under specific hematology/MP sections rather than miscellaneous.
- NLP was applied to make free-text information searchable.
- An interface was established between the ordering lab and reference lab for electronic exchange.
- Synoptic structures were tested to ensure discrete data fields (positive/negative status, comments) were sent electronically.
Answer The solution replaced the miscellaneous PDF attachment with discrete fields in the EMR lab flowsheet, allowing clinicians to view results without extra clicks and enabling automated trending and research use.
Common mistakes
- Storing lab results as PDF attachments in generic categories like 'miscellaneous', which prevents automated searching and trending.
- Failing to establish interfaces between the LIS and external reference labs, leading to manual data entry and errors.
- Neglecting to engage stakeholders from all affected areas, such as revenue teams, resulting in incomplete solutions.
- Not standardizing data formats through synoptic reporting, which hinders integration from disparate sources.