Site-of-care steering programs — insurer policies that try to move an infusion from a physician's office to a lower-cost setting, or vice versa — are usually justified by a simple claim: hospitals get paid dramatically more than offices for the same drug. The raw numbers back that up. Once you control for how the claim is billed, most of the gap disappears.
| Comparison | Gap vs. office | What it measures |
|---|---|---|
| As-billed | +32.1% | Every hospital-outpatient claim vs. every office claim for the same drug, billing form not held constant. |
| Within billing class | +6.0% | Same billing form compared on both sides, matched payer × state × drug. |
As billed, hospital outpatient departments appear to collect about 1.321× what an office collects for the identical drug — a 32% premium, the number usually quoted in the site-of-care debate. Once the comparison is restricted to matched pairs on the same billing form, the gap falls to 1.06×, about a 6% premium.
Office (place of service 11) and hospital outpatient (place of service 22) claims are not filed on the same form and are not priced the same way even for an identical drug and dose. Hospital outpatient claims run through an institutional billing process with its own fee schedule logic; office claims run through a professional fee schedule. Comparing the two as-billed mixes the site-of-care question with a billing-form question, and the billing-form difference does most of the work: it is not that the hospital is being paid dramatically more to administer the same drug — it is that the two settings are not filing the same kind of claim in the first place. When the comparison is restricted to the same billing form on both sides, matched payer by payer, state by state, and drug by drug, most of the premium disappears.
The within-billing-class comparison rests on n = 1,099 matched payer × state × drug pairs — stated here plainly as a small matched set, not a population-scale result. It is large enough to move the headline figure from a 32% gap to a 6% gap, which is itself the finding worth publishing, but it is not large enough to support drug-by-drug or state-by-state breakdowns with any confidence, and it has not moved across a 4× expansion of the underlying corpus in this vintage.
Contracted rates from insurers’ machine-readable files, not amounts paid on a claim. Figures are drawn from report_data.json published alongside the Infusion Reimbursement Index and are not independently re-derived on this page.