Selective direct bilirubin testing in outpatient liver chemistry: temporal validation and decision curve analysis

Total bilirubin (TBil) and direct bilirubin (DBil) are frequently ordered together in outpatient liver chemistry testing, although routine paired ordering may generate low-value DBil measurements. We evaluated predefined selective DBil testing policies as a pre-implementation laboratory-utilization analysis. Adult outpatient liver chemistry records from 2024 and 2025 were analyzed retrospectively. The 2024 cohort was used for strategy evaluation, and the 2025 cohort was used for temporal validation. Test-all, TBil-only, and TBil/ALP/GGT strategies were compared for DBil >8 μmol/L and DBil >16 μmol/L. Additional analyses varied the TBil trigger threshold, examined ALT-inclusive and sex-specific threshold variants, assessed patient- and department-level robustness, and descriptively reviewed records missed by the selective strategies. We analyzed 22,455 records from 2024 and 20,985 from 2025. In the 2025 validation cohort, TBil-only avoided 91.24% of DBil tests and missed 11 DBil >8 μmol/L results (8.1–10.7 μmol/L), whereas TBil/ALP/GGT avoided 71.03% and missed one result (8.1 μmol/L). Neither main reduced-testing strategy missed DBil >16 μmol/L. Lowering the TBil trigger from 23 to 20 μmol/L reduced TBil-only misses from 11 to 4 while retaining 85.70% test avoidance. Decision curve analysis favored TBil-only, whereas TBil/ALP/GGT occupied a higher sensitivity, lower-reduction position. These retrospective data quantify the trade-offs among locally defined selective DBil testing policies but do not establish patient-level clinical safety or a universal threshold. TBil-only was the efficiency-oriented option, whereas TBil/ALP/GGT was more conservative for detecting mild DBil elevations. Prospective pilot implementation, clinician over-riding, local threshold validation, and post-implementation audits are required before routine adoption.