Nearly Half of Medicare’s 31 Drug-Price Flags Recurred in at Least Five Quarters

The problem is persistence, not simply a one-quarter anomaly

Potential errors in manufacturer-reported drug prices kept 31 Medicare Part B drug codes out of a federal price-substitution assessment based on 2024 data. What’s the Scoop With Broach’s analysis finds that nearly half of that group—15 of 31, or 48.4%—had potential errors flagged in at least five of the preceding 12 quarters. The annual inspector-general report’s count is therefore not merely a snapshot of newly questionable data.

The September 21, 2026 review says 17 of the 31 codes had potential errors reported to CMS in at least three quarters of 2024. Of those 17, 15 also had potential errors in at least five quarters of the 2022–2024 lookback. The latter group represents 88.2% of the repeatedly flagged 2024 subgroup. These are nested counts: adding 17 and 15 would count most of the same drugs twice.

The public-interest question is whether the pricing safeguard can reliably act on persistent data problems while avoiding payment cuts based on incorrect information. The documents establish a repeated barrier to assessment, not 31 proven missed price cuts or a calculable total of money lost. The report also credits CMS with correcting potential errors during 2025, an important qualification to a claim that the agency took no action.

A price difference is not the same as eligibility for a cut

The safeguard compares average sales price with average manufacturer price. Under CMS’s policy, a drug generally must exceed the 5% comparison threshold in two consecutive quarters or three of the previous four quarters, using complete comparable product data. A substitution cannot raise the payment, and drugs identified as in short supply are excluded. Those conditions protect both the program’s finances and patient access.

Only seven of the 31 potentially erroneous drug codes crossed the 5% threshold in at least one quarter—22.6% of that group. Crossing once does not prove that the drug satisfied the policy’s repeated-quarter test, the complete-data condition or the other safeguards. It would therefore be unsupported to call all seven confirmed candidates, much less all 31.

We checked that distinction against the separate fourth-quarter 2024 pricing memorandum, dated May 12, 2025. It identified eight drug codes meeting the price-substitution criteria and another eight exceeding 5% in that quarter but not meeting other criteria. Thus even with complete price data, half of the 16 threshold-crossing codes in that quarterly comparison were not eligible under the policy. This is a separate quarterly population, not a subdivision of the annual 31-code error group.

Fourteen drugs received substitutions; two lacked a savings calculation

The annual report lists 14 codes for which CMS implemented substitutions based on 2024 price data. Counting its appendix entries shows 21 drug-quarter substitutions across the fourth quarter of 2024 through the third quarter of 2025: five in each of the first three displayed quarters and six in the last. Palonosetron, code J2469, appears in all four quarters. Repeat quarters are not additional distinct drugs.

The report attributes $1.6 million in one-year savings to the 2024-data substitutions and $78 million cumulatively since 2013. But endnote 4 says savings could not be calculated for two of the 14 drugs because their codes changed and sales-price data were unavailable for the implementation quarter. The published dollar figure is therefore not a complete per-drug calculation for every code receiving a substitution.

The review’s methodology subtracts the substituted manufacturer-price-based payment from the sales-price-based payment after sequestration adjustments, then multiplies by utilization in the implementation quarters. We did not recreate that calculation from confidential manufacturer prices. Nor do we divide the aggregate savings by 14 and present an equal saving for each drug: utilization, prices, substitution duration and the two missing calculations differ.

CMS corrections and the limit of the public record

The inspector general says CMS reviewed and corrected potential errors throughout 2025, sometimes working with manufacturers. Separately, four codes had potential sales-price errors; CMS explained that the error did not affect payment calculations or corrected it. Those four are not the same count as the 31 codes with potential manufacturer-price unit or amount errors, and we do not automatically add them to form a distinct-drug total.

The reason for withholding a price recommendation is also consequential: a faulty unit type or price amount could cause an inappropriate reimbursement cut and threaten access. The annual report encourages continued correction work but issues no formal recommendations. It does not name the 31 excluded codes, quantify foregone savings or determine whether manufacturers subsequently updated their data.

The identified safeguard limitation is limited but substantive: recurrent uncertainty prevented assessment of some drugs, even while verified substitutions produced measurable savings and some errors were corrected. Neither a claim of zero progress nor a claim that every flagged record has now been resolved is supported by these documents.

Method

This original document analysis compares the annual report’s nested error counts, appendix quarter markers, savings methodology and endnotes with the separate fourth-quarter pricing memorandum. We calculated the persistence shares, counted drug-quarter substitutions and distinguished single-quarter price differences from full policy eligibility. We did not conduct firsthand pharmacy reporting, test manufacturer submissions or make a clinical recommendation.

As of October 14, 2026, the reviewed evidence does not establish the present correction status of every flagged drug or an amount of recoverable overpayments. The underlying reporting year is 2024 and the implemented-payment window extends into 2025. Keeping those dates separate is essential: these records do not show how often the same problem occurred in 2026.