Researchers have developed Binary PheNorm, an extension of the PheNorm algorithm designed to more effectively utilize binary silver labels from electronic health records (EHRs) for computable phenotyping. Unlike the original PheNorm, which is suited for count-valued labels, Binary PheNorm directly incorporates binary indicators, such as NLP mentions or medication flags, into its denoising process. This approach yields a continuous phenotype score without requiring Expectation-Maximization calibration and can be combined with count-based labels for improved performance. In clinical simulations, Binary PheNorm demonstrated significant improvements in discrimination for conditions like anaphylaxis and acute pancreatitis, outperforming simple indicator-based methods. AI
IMPACT Enhances the utility of electronic health records for research by improving the accuracy of phenotype extraction from binary data.
RANK_REASON The cluster contains an academic paper detailing a new methodology for data analysis in the field of computational phenotyping. [lever_c_demoted from research: ic=1 ai=0.7]
- acute pancreatitis
- anaphylaxis
- Binary PheNorm
- electronic health records
- natural language processing
- PheNorm
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
