Verification — run the implementation against every value published in the source papers
These tests call the same functions the chart above uses, so the chart cannot
pass while the model is wrong. Each row cites where its expected value comes from. Tolerances are
stated per test; crossover hours are compared to figures read to the nearest hour in the source,
so they carry a ±1 h tolerance.
Thresholds implement the consensus line segments of Stavis RL, Guida JD. J Perinatol 2026
(doi:10.1038/s41372-026-02855-y), interpolated by day of postmenstrual age, with the AAP 2022
phototherapy thresholds for 35 weeks with a neurotoxicity risk factor (Kemper AR et al. Pediatrics
2022;150:e2022058859, Supplemental Table 2) substituted at any hour where they are lower. The AAP
table has no hour-0 entry; the hour-1 value is used at birth. Lines run to 35 0/7 weeks' postmenstrual
age for infants born at or below 33 0/7 weeks, and to 14 days of age for infants born later.
Draft evaluation tool — not validated for clinical use, and not an Epic product.