Where the cost inputs for a Brazilian health economic model actually come from
The sources are public and easy to name, and that is the smallest part of the job, because the records they hold are not usable as they stand.
A team building a model outside Brazil can find the databases in an afternoon, and then stall, because a keyword search of public procurement misses real products, merges unrelated ones, and prices a premium item at the cost of a commodity. What turns those open records into a defensible input is the reconstruction in between, and this is a map of both, the sources that answer each input, and the work that makes them usable.
Why the raw data is not usable
Brazilian government procurement is open, and that openness is deceptive. The same product is recorded under different catalogue categories, so a single catalogue code can cover both a specialised product and a commodity substitute bought at a fraction of the price, so the technical specification alone cannot tell them apart. Suppliers often enter the importer's name, a reference code or a generic term instead of the brand, so one generic descriptor stands in for the products of several different manufacturers. Prices sit against units, boxes and packs of different sizes, across several years. A keyword search across that will miss products and merge others, and any average built on top inherits the mess.
Product and device prices: the Banco de Preços em Saúde
The Banco de Preços em Saúde (BPS), maintained by the Ministry of Health, records what public institutions actually paid, purchase by purchase, with brand, supplier, quantity, unit and buying institution, and it is where a defensible product cost begins. Reaching a usable figure from it means reading every relevant category rather than the obvious ones, which for a single product family can run to tens of thousands of purchase records, and identifying each product through the ANVISA registry, which returns the commercial names, the registration holder and the manufacturer, and so recovers products sold under an importer or sister-product name and flags those registered for a different use than expected.
From there the classification has to follow the brand and not the catalogue code, so that a premium product is counted only where the brand is recorded and is never priced at the cost of a commodity substitute, and every purchase is made comparable, updated to a single reference date with the national price index, converted to a common unit, screened for outliers, and reported with the number of purchases behind each figure. Where a product has no public purchase at all, the honest output names an equivalent of the same technical specification and flags it as such.
Medicines: the BPS price against the CMED ceiling
For medicines there are two public anchors, the price actually paid, again from the BPS, and the government ceiling price set by CMED, the PMVG, with the paid price usually the better base case and the ceiling the complement when the purchases are too few to be reliable. Before either, the ANVISA registry settles whether the product can be bought at all, which is not a formality, since a requested molecule can have no active registration, in which case the base case moves to an alternative used in Brazilian practice rather than to a price that does not exist.
Resource use: SIGTAP is reimbursement, not cost
For a consultation, a procedure or a hospital day, SIGTAP gives the amount the public system reimburses, which is the correct figure from the payer's side and, by design, far below the cost of delivering the care, since SIGTAP pays on the order of ten reais for a specialist consultation. A real-cost estimate has to be built rather than looked up, with hourly staff costs derived from national public-sector payroll across every public employer, including employer contributions and productive hours, combined with the published Brazilian costing studies that exist for some items and not for others, so that each parameter carries the SIGTAP value for reference, a real-cost base case, and a range wide enough for the sensitivity analysis where no Brazilian study measures the item directly.
Why the inputs are what survives review
When a dossier reaches CONITEC, REBRATS or the ANS, the cost inputs are among the first things examined, and critical reviews of submitted analyses point repeatedly to weak direct-cost evaluation and outdated sources as recurring faults. An input that names its source, states its window, declares its summarising statistic and shows the purchases behind it is the one that holds up under that reading, and the methods note that carries all of it is what makes the number auditable by the counterparty who has to defend it.
When a model built elsewhere needs Brazilian inputs
corina turns Brazilian public data into model-ready inputs, reading every relevant category, reconstructing each product through the ANVISA registry, classifying by brand, normalising prices to one reference date and a common unit, and building real resource-use costs from public-sector payroll rather than SIGTAP reimbursement, delivered in your template with a purchase-level annex and a methods note the counterparty can audit, as a single input, a Brazil workstream inside a global programme, or ongoing specialist support.