Data Consult Service: Can we use observational data to address immediate clinical needs?

Area of Science:

• Clinical Informatics
• Observational Data Research
• Real-time Evidence Generation

Background:

• Clinical decision support tools often struggle with biases in observational data.
• There is a need for methods to generate timely clinical evidence from real-world data.

Purpose of the Study:

• To describe the operational experience of a data consult service generating real-time clinical evidence.
• To characterize challenges encountered when using observational data for clinical decision support.

Main Methods:

• A data consult service pipeline was implemented, including question gathering, data exploration, patient phenotyping, study execution, and validity assessment.
• User feedback was collected to evaluate the process and identify issues.
• 29 clinical questions were gathered from 22 clinicians via various communication methods.

Main Results:

• The service successfully answered 24 out of 29 clinical questions.
• Different question types emerged based on data collection methods (e.g., clinical rounds yielded more patient characterization queries).
• Key challenges included missing/incomplete data, underreported conditions, nonspecific coding, and accurate drug regimen identification.

Conclusions:

• The Data Consult Service shows potential for evidence generation but faces limitations in real-time question answering.
• Addressing challenges in patient phenotyping and study design is crucial.
• Adherence to validated practices for observational research is mandatory for producing reliable clinical evidence.