A case study evaluating the portability of an executable computable phenotype algorithm across multiple institutions...

Area of Science:

• Health Informatics
• Computational Biology
• Clinical Research Informatics

Background:

• Electronic health record (EHR) algorithms for patient cohort identification are often shared as free-text, necessitating manual interpretation and implementation.
• This manual process introduces variability and inefficiency in clinical research and data analysis.

Purpose of the Study:

• To develop and evaluate a standardized, computable phenotype algorithm architecture (PhEMA) for automating patient cohort definition.
• To assess the feasibility and performance of PhEMA in executing a benign prostatic hyperplasia (BPH) algorithm across multiple healthcare sites.

Main Methods:

• Developed the Phenotype Execution and Modeling Architecture (PhEMA) to author and execute standardized computable phenotype algorithms.
• Converted a BPH algorithm from the electronic Medical Records and Genomics (eMERGE) network into a PhEMA-compatible format.
• Deployed the computable algorithm to eight sites for execution against local data warehouses and i2b2 instances.

Main Results:

• Six out of eight participating sites successfully executed the computable BPH algorithm.
• Positive predictive values (PPV) for identified cases were ≥90% based on blinded chart review.
• High overlap (>90%) in selected cases was observed between the computable algorithm and original implementations at 3 out of 5 sites.

Conclusions:

• PhEMA demonstrates the potential to automate phenotyping across diverse EHR systems using standardized computable representations.
• Successful execution and high PPV indicate the utility of PhEMA for reproducible clinical research.
• Ongoing challenges in implementation across different data infrastructures require further attention.