Abstract: SA-PO0081
Diagnostic Aggregation and Semantic Drift in International Classification of Diseases (ICD)-Derived Kidney Phenotypes: Implications for Electronic Health Record (EHR)-Based and Genomic Research
Session Information
- Artificial Intelligence: Kidney Imaging, Pathology, and Precision Phenotyping
October 24, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Artificial Intelligence, Digital Health, and Data Science
- 300 Artificial Intelligence, Digital Health, and Data Science
Authors
- Nestor, Jordan Gabriela, Columbia University, New York, New York, United States
- Krishna Murthy, Sarath Babu, Columbia University, New York, New York, United States
- Zachariah, Teena, Columbia University, New York, New York, United States
- Navarro Torres, Mariela, Columbia University, New York, New York, United States
- Khan, Atlas, Columbia University, New York, New York, United States
- Radhakrishnan, Jai, Columbia University, New York, New York, United States
- Kiryluk, Krzysztof, Columbia University, New York, New York, United States
- Weng, Chunhua, Columbia University, New York, New York, United States
Background
Kidney phenotypes derived from ICD codes are frequently re-represented across research terminologies such as Phecodes, introducing additional abstraction. Whether these transformations preserve diagnostic distinctions and traceability to source diagnoses remains unclear.
Methods
We identified kidney-relevant ICD-9-CM and ICD-10-CM codes using clinician-curated OHDSI ATLAS searches and evaluated real-world usage in a large academic health system (n=571,435). ICD codes were mapped to Phecodes, SNOMED-CT, Human Phenotype Ontology (HPO), and Online Mendelian Inheritance in Man (OMIM). Mapping structure was characterized by coverage, mapping richness, and multiplicity, with multiplicity operationalizing diagnostic aggregation. Three blinded nephrologists rated semantic fidelity (good, fair, poor), and ordinal logistic regression assessed its association with mapping multiplicity.
Results
Among 585 kidney-relevant ICD codes, mapping coverage and structure varied across terminologies. Phecodes showed high coverage but high mapping multiplicity, whereas SNOMED-CT, HPO, and OMIM mappings were predominantly one-to-one. Semantic fidelity differed substantially: 39% of ICD-Phecode mappings were rated poor, while no poor matches were observed for the other terminologies. Higher mapping multiplicity was associated with lower semantic fidelity (OR per log increase, 0.33; 95% CI, 0.29-0.37; p<0.001). Domains such as Glomerulopathy and Chronic kidney disease showed the highest rates of diagnostic aggregation and semantic distortion.
Conclusion
Diagnostic aggregation is a measurable mechanism through which ICD-derived kidney phenotypes lose semantic fidelity and traceability to source diagnoses, representing a potential source of misclassification in genomic studies and limiting identification of clinically meaningful subgroups. Careful phenotype construction is essential for ensuring kidney genomics findings remain interpretable and translatable to patient care.
Acknowledgment
This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant K08DK132511 (JGN) and the American Society of Nephrology (ASN) and Kidney Cure Harold Amos Medical Faculty Development Award (JGN). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding entities.
Examples of Diagnostic Distortion in ICD-Derived Kidney Phenotypes
| ICD Code | ICD Label | Terminology | Mapped Concept | Semantic Fidelity Outcome | Type of Distortion |
| ICD-9-CM: 753.16 | Medullary cystic kidney | Phecode | Cystic kidney disease (751.21) | Failed | Loses disease-level specificity |
| ICD-10-CM: N00.1 | Acute nephritic syndrome with focal and segmental glomerular lesions | Phecode | Proliferative glomerulonephritis (580.11) | Failed | Loses temporal meaning and histopathologic subtype |
| ICD-10-CM: N02.8 | Recurrent hematuria with morphologic changes | Phecode | Nephrotic syndrome (580.2) | Failed | Incorrect disease context introduced |
| ICD-9-CM: 753.11 | Congenital single renal cyst | SNOMED-CT | Congenital single renal cyst (5941000119101) | Preserved | One-to-one mapping |
| ICD-10-CM: Q60.3 | Renal hypoplasia, unilateral | HPO | Unilateral renal hypoplasia (HP:0012583) | Preserved | Disease-level specificity and laterality retained |
| ICD-9-CM: 753.16 | Medullary cystic kidney | OMIM | Medullary cystic kidney (MTHU018957) | Preserved | One-to-one mapping |
Funding
- NIDDK Support