Researchers analyzed PEDSnet, a network of pediatric hospitals that share electronic health records EHRs data for research purposes, to determine the positive predictive value (PPV) of diagnostic codes for screening patients for primary hyperoxaluria (PH). 

Their study was a cross-sectional analysis of data collected from children at 7 PEDSnet facilities between 2009 and 2021. Using diagnostic codes, they constructed 3 groups of hypothesized PH probabilities, which they used to inform the development and implementation of screening criteria. Codes for PH were particular to the first tier, whereas codes for hyperoxaluria, oxalate nephropathy, and oxalosis were specific to the second tier, and codes more than or equal to 2 for disturbance of carbohydrate metabolism and more than or equal to 1 for kidney stones were combined for the third tier. They analyzed the electronic health records EHRs of patients suspected of having PH to verify the diagnosis, as well as to gauge how precise and timely the diagnostic codes were. 

Tiers, times, PH types, and locations were analyzed to see which factors had the greatest impact on the PPV of the codes. After screening 341 people, they found that 33 (9.7%) had definite PH. Tier 1 had the highest percentage of PH, although it only had a 20% PPV. Higher PPV was predicted by a hospital’s ability to correctly portray point-of-care diagnoses during data extraction. PH3 had a perfect PPV (100%) for its diagnostic code, whereas PH1 had the lowest (22.8%). The PPV of PH diagnostic codes is low. The results point to the need for caution when conducting research using huge databases when source validation is not an option.