← Research programmes
Agent · health system economicsPrototype
A compass for hospital medicines spending
Status. Research use. Management analytics. No patient-level output, no prescribing access, and no recommendation about any medicine for any patient.
The setting
- An Italian health authority already holds every figure it needs to see that a biosimilar has been available for eight months and it is still buying the originator. It sees it the following year, in a report.
- The problem is not missing data. It is that the data is not organised in a way that allows a decision while the financial year is still open.
What we did
Built a read-only analytical layer over the pharmaceutical flows an authority already produces for other statutory obligations. Reconcile the totals against the balance sheet first, then normalise package prices to a comparable unit, then report.
The tool does not generate the saving. A patent expiry generates the saving from loss of exclusivity, a clinician and a pharmacist generate the saving from a switch, a negotiated agreement generates a recovery. What a governance tool can defensibly cause is one thing only: a reduction in the delay with which the organisation reacts to an opportunity that already exists.
What exists now
- A working prototype, built and run against real expenditure records from Italian health authorities.
- No figures are published here. The record counts and expenditure totals sit in a commercial document, they are marked confidential there, and none of them is recomputable by a reader. The method is what is publishable, and it is the stronger asset anyway.
What it does not do
- It has no access to prescribing systems and produces no patient-level output. The minimum unit of aggregation is the dispensing structure.
- It makes no recommendation about any medicine for any patient. Clinical substitutability stays entirely with the pharmacist and the clinician.
- Biosimilar switching is a clinical decision, not a procurement one. If the tool ever touched switching, formulary preference or therapeutic substitution directly, the non-device framing would not hold.
What has not been established
- Whether any authority actually acted faster.
- The reaction-time indicator is the thing that would prove the method, and it has not been measured on anybody.
Where it goes next
- The attribution principle above is a testable hypothesis and it has not been tested: that making an already-existing opportunity visible shortens the time an organisation takes to act on it.
- That is a measurable claim about organisational behaviour, with a clean before-and-after design and an indicator a third party can verify.
What we are looking for
Research partners to test the hypothesis. This is a study, not a deployment: what it needs is a data-sharing agreement and an agreed baseline, not a procurement process.
Get in touch →