Assessing ICD-10-CM Algorithms for Repeated eGFR-Defined Chronic Kidney Disease in the TriNetX Network

By
Chung-An Wang
Wei-Jie Wang
Jui-Yi Chen
Chung-Yi Cheng
Vin-Cent Wu
October 13, 2026

Bmj Health & Care Informatics

Overview

In 8,239,691 adults in the TriNetX US Collaborative Network with at least two eGFR measurements, 1,328,952 met laboratory criteria for CKD stage ≥3. ICD-10-CM algorithms showed high specificity and negative predictive value but moderate sensitivity; requiring repeated codes or restricting code ascertainment around the index eGFR increased PPV while reducing sensitivity.

Background

CKD ascertainment in electronic health records supports population research and longitudinal evaluation, but diagnosis codes may not capture all patients with laboratory-defined disease. The source notes prior evidence of under-recognition, including a TriNetX REVEAL-CKD finding that 64.3% of US patients with stage 3 CKD had no recorded CKD diagnosis. This study evaluated ICD-10-CM algorithms against repeated eGFR-defined CKD in the TriNetX network. KDIGO guidance defines CKD by kidney structure or function abnormalities present for at least 3 months and cautions against inferring chronicity from a single abnormal result.

Data Highlights

Population or algorithm	Sensitivity	Specificity	PPV	NPV
Adults with ≥2 eGFR measurements, 2021–2025	8,239,691
Laboratory-defined CKD stage ≥3	1,328,952
Stage ≥3; ≥1 ICD-10-CM code	64.1%	95.6%	73.8%	93.3%
Stage ≥4; ≥1 ICD-10-CM code	79.2%	98.0%	57.3%	99.3%
Temporal restriction around index eGFR (±3 to ±12 months)	Declined	Not reported	Approximately 96% for stage ≥3; 93% for stage ≥4	Not reported

Key Findings

The analysis included 8,239,691 adults with at least two eGFR measurements from 2021 through 2025; 1,328,952 met laboratory-defined criteria for CKD stage ≥3.
For stage ≥3, requiring at least one ICD-10-CM code yielded 64.1% sensitivity, 95.6% specificity, 73.8% PPV, and 93.3% NPV.
For stage ≥4, at least one code yielded 79.2% sensitivity, 98.0% specificity, 57.3% PPV, and 99.3% NPV.
Requiring at least two ICD-10-CM codes decreased sensitivity and increased PPV.
Restricting code ascertainment to ±3, ±6, or ±12 months of the index eGFR raised PPV to approximately 96% for stage ≥3 and 93% for stage ≥4, while sensitivity declined.
The source reports that 64.3% of US patients with stage 3 CKD in the TriNetX REVEAL-CKD study had no recorded CKD diagnosis.

Clinical Implications

For research using TriNetX, the choice of ICD-10-CM phenotype should reflect whether the application prioritizes diagnostic certainty or case ascertainment, because repeated coding and temporal restrictions increased PPV while reducing sensitivity. The study concludes that no single ICD-based phenotype or temporal strategy is universally optimal and notes that combining diagnosis codes with longitudinal laboratory data may improve identification.

Conclusion

In the TriNetX network, ICD-10-CM algorithms identified repeated eGFR-defined CKD with high specificity but incompletely captured cases. Their performance varied with code-count and timing requirements, producing a trade-off between PPV and sensitivity.