Diagnostic Code-Based Screening for Identifying Children with Primary Hyperoxaluria. Issue 4 (5th October 2022)

• Record Type: Journal Article
• Authors: Tasian, Gregory; Dickinson, Kimberley; Karafilidis, John; Marchesani, Nicole; Antunes, Nuno; Razzaghi, Hanieh; Utidjian, Levon; Yonekawa, Karyn; Coplen, Doug; Muneeruddin, Samina; DeFoor, Bob; Rove, Kyle O.; Forrest, Christopher; Ching, Christina
• Is Part Of: Journal of urology. Volume 208:Issue 4(2022)
• Page Start: 898
• Page End: 905
• Publication Date: 2022-10-05
• Subjects: hyperoxaluria, primary -- electronic health records -- validation study; Genitourinary organs -- Periodicals; Urology -- Periodicals; Urology -- Periodicals; Urologie -- Périodiques; Urologie; 616.6
• DOI: 10.1097/JU.0000000000002863
• Languages: English
• ISSNs: 0022-5347

Abstract

Abstract : Purpose: We evaluated the utility of diagnostic codes to screen for patients with primary hyperoxaluria (PH) and evaluate their positive predictive value (PPV) in identifying children with this rare condition in PEDSnet, a clinical research network of pediatric health systems that shares electronic health records data. Materials and Methods: We conducted a cross-sectional study of children who received care at 7 PEDSnet institutions from January 2009 through January 2021. We developed and applied screening criteria using diagnostic codes that generated 3 categories of the hypothesized probability of PH. Tier 1 had specific diagnostic codes for PH; tier 2 had codes for hyperoxaluria, oxalate nephropathy, or oxalosis; and tier 3 had a combination of ≥2 codes for disorder of carbohydrate metabolism and ≥1 code for kidney stones. We reviewed the electronic health records of patients with possible PH to confirm PH diagnosis and evaluate the accuracy and timing of diagnostic codes. The PPV of the codes was compared across tiers, time, PH type, and site. Results: We identified 341 patients in the screen; 33 had confirmed PH (9.7%). Tier 1 had the highest proportion of PH; however, the PPV was only 20%. The degree to which an institution accurately represented point of care diagnoses in the data extraction process was predictive of higher PPV. The PPV of diagnostic codes was highest for PH3 (100%) and lowest for PH1 (22.8%). Conclusions: Diagnostic codes for PH have poor PPV. Findings suggest that one should be careful in research using large databases in which source validation is not possible.