Same Top Fraction, Different Workload: Auditing Eligibility and Capacity in Clinical Risk Evaluation

Clinical risk scores turn patient rankings into finite worklists for specialist review, monitoring, or prevention. Comparing the same top fraction across candidate sets of different sizes changes both who can be ranked and how many people receive service. We introduce a decision-aware four-cell audit that keeps the full population event denominator fixed and crosses two eligibility rules with two absolute budgets, separating broader reach from additional capacity. We apply it to frozen birth-record score vectors for preterm birth, neonatal intensive care admission, and low birth weight in approximately 3.48 million CDC Natality 2023 records. At the nominal top 10% operating point, moving from early-entry eligibility to a retrospective all-record upper bound reports a 39–42% relative increase in captured events while expanding the worklist by one third. Symmetric accounting allocates about three fifths of the joint gain to added capacity and two fifths to broader eligibility; at a fixed budget, the eligibility gain is less than half the cohort relative estimate. A complete replication on CDC 2024 preserves this allocation pattern. In 18 of 81 controlled settings, the mean cohort-relative contrast is positive while fixed-budget capture declines. The audit therefore distinguishes population reach from resource expansion and makes the policy represented by a top fraction metric explicit. Code, trained models, and experiment results are available at https://github.com/vache4ogan/prenatal-risk-scoreability.