Industries / Health / Healthcare / Plasma and blood centers
Testing
Testing is the work of checking each donation's blood group, its match with the patient and its freedom from infection before it is released to hospitals.
7
Milestones achieved, from 1901 to 2011.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1901
Blood groups discovered
Early transfusions between people often ended in disaster because mixed blood sometimes clumped. Karl Landsteiner explained in 1901 that people have different types of red blood cells, called blood groups, and the Nobel Prize's own summary says the discovery made safe transfusion between compatible groups possible.
Credit
Karl Landsteiner, Nobel Prize in Physiology or Medicine, 1930
1945
The antiglobulin test
Robin Coombs, Arthur Mourant and Robert Race published "a new test for the detection of weak and 'incomplete' Rh agglutinins", antibodies against the Rh blood group that Landsteiner and Alexander Wiener had reported in 1940 and 1941. The test, often called the Coombs test, is used before transfusion to check a patient's blood for antibodies against donor red cells.
Credit
R. R. A. Coombs, A. E. Mourant and R. R. Race
1985
First HIV test for donations
On 2 March 1985 the US Food and Drug Administration (FDA) licensed the first commercial blood test for HIV, and blood banks began screening the US supply, by the US government's HIV.gov timeline. Abbott, which developed the test, cites an estimate that before screening about 1 in 90 transfusions in some US cities carried HIV. By WHO's figures for 2023, 10 reporting countries still could not screen all donated blood for HIV, hepatitis B, hepatitis C and syphilis.
Credit
Abbott; the US FDA
1996
UK starts tracking transfusion harm
Serious Hazards of Transfusion (SHOT) began collecting anonymised reports of harm and near misses from every UK organisation that transfuses blood, work called haemovigilance. In its 2025 report, 83.0% of 4,046 reports were errors, no confirmed or probable transfusion-transmitted infection was reported, and SHOT put the risk of a transfusion-related death at about 1 in 40,000 components issued.
Lessononce screening made infection rare, most reports concerned mistakes in handling and checking, not the blood itself.
Credit
SHOT, funded by the UK blood services
1997
Krever report on tainted blood
On 25 November 1997 Justice Horace Krever reported on how Canada's blood system handled HIV and hepatitis C: about 1,000 people transfused from the late 1970s to the 1980s were infected with HIV and another 30,000 with hepatitis C. He criticised delays, including waiting for a randomised trial before adopting surrogate testing for hepatitis C, that is, screening for indirect signs of infection, and his 50 recommendations led to Canadian Blood Services and Héma-Québec replacing the Canadian Red Cross.
Lessonin Krever's words, "action to reduce risk should not await scientific certainty.".
Credit
Commission of Inquiry on the Blood System in Canada, led by Justice Horace Krever
1999
Nucleic acid testing begins
In mid-1999 US blood centres added nucleic acid testing (NAT), which looks for a virus's genetic material rather than the antibodies the body makes weeks later, for HIV and hepatitis C. In its first three years it found 12 HIV-infected donations among 37.2 million, and 170 hepatitis C-infected donations among 39.7 million, that antibody tests had passed, and the study team put the remaining risk of each at about 1 in 2 million units. The UK's advisory committee, JPAC, estimated for 2022 to 2024 that testing missed about 0.10 potentially infectious HIV donations per million.
Credit
the National Heart, Lung, and Blood Institute NAT Study Group (Susan Stramer, Michael Busch and colleagues)
2011
Swiss treat every platelet unit
Platelets are stored warm, and the Swiss authors call bacterial contamination of them the most common infection risk of transfusion. In 2011 Switzerland began treating all platelets with INTERCEPT pathogen inactivation, a process meant to disable germs that screening misses. National haemovigilance data showed 16 transfusion-transmitted bacterial infections, 3 of them fatal, among 158,502 untreated units issued from 2005 to 2011, and none among 205,574 treated units from 2011 to 2016.
Credit
the Swiss Red Cross blood transfusion services and Swissmedic, the Swiss regulator; Cerus Corporation makes INTERCEPT
Watch for
Pathogen-reduced red cells
Cerus says its Phase 3 RedeS trial met its main goal: patients given INTERCEPT-treated red cells had an average haemoglobin rise (haemoglobin is the protein that carries oxygen) of 1.996 g/dL, against 2.050 g/dL for standard red cells, in 195 and 199 patients, a difference inside the trial's preset margin. Six of 267 treated patients (2.2%) formed antibodies specific to the treated cells, against none of 276 in the control group. The company states the red-cell system is not approved by the FDA.
Achieved whenthe full results appear in a peer-reviewed journal, or the FDA or a European notified body approves the system for red cells.
Claimed by
Cerus Corporation, November 2026
Open
01
Harm figures after ISTARE
A free, machine-readable table of the yearly totals that national haemovigilance schemes publish, from 2016 on: components issued, reactions in patients by category, complications in donors by category, deaths and confirmed transfusion-transmitted infections. It covers at least 12 countries, links every number to the report page it came from, flags where a scheme counts differently, and recomputes rates on one stated denominator wherever the report gives the counts. Published national totals only, with no case records and no donor-level data.
The briefCredit · unclaimed02
Testing rules as code
The rules for which tests a donation must pass before release, written as machine-readable decision tables (tables that give an answer for each combination of conditions) for two rule books: the US rule, 21 CFR 610.40, and the UK blood services' testing rules. With them goes a public set of at least 300 synthetic donation cases, each with the correct "release" or "hold" answer and the clause it tests, plus a script that runs any blood-bank program against the set and scores it.
The briefCredit · unclaimed
Open 01 · the brief
Harm figures after ISTARE
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free, machine-readable table of the yearly totals that national haemovigilance schemes publish, from 2016 on: components issued, reactions in patients by category, complications in donors by category, deaths and confirmed transfusion-transmitted infections. It covers at least 12 countries, links every number to the report page it came from, flags where a scheme counts differently, and recomputes rates on one stated denominator wherever the report gives the counts. Published national totals only, with no case records and no donor-level data.
Why it is still open
The only worldwide set, the International Haemovigilance Network's ISTARE database, holds 125 national data sets from 25 countries for 2006 to 2016 only, is released on request by email, and was limited, the network says, by inconsistent use of definitions. The closest current work, the European Commission's yearly SARE report, is a 77-page PDF on 30 European countries that covers patients, not donors, and notes that SHOT counts per 10,000 units issued while the EU, the FDA and Canada count per 100,000 transfused. The largest US plasma donor study used industry definitions. A GitHub search found reporting-form apps but no compiled dataset.
Built with what exists
The schemes' public annual reports, starting with SHOT's from 1996–97 to 2025; pdfplumber or Camelot to pull tables out of PDFs; a long-context AI model that proposes each value with its page number for a person to accept, in the manner of the source-citing reviewer agent (Note: "Findings that cite notes"); the ISBT donor-complication standard and the haemovigilance definitions that WHO, the ISBT and the IHN made free to all (Vox Sanguinis, 2024).
First step
Enter SHOT's yearly totals for 2016 to 2025 from its annual summaries, then list which of the schemes in the SARE 2024 benchmarking table (Canada, Australia, the US, South Africa, New Zealand) publish yearly figures, and which print donor reactions, online.
Done looks like
One CSV under an open licence with yearly rows for 2016 to 2024 from at least 12 national schemes, at least 6 of them outside the SARE report and at least 5 with donor figures. Every value carries its report link and page, and a second person checks 200 randomly chosen values with at least 98% matching the source.
Open 02 · the brief
Testing rules as code
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
The rules for which tests a donation must pass before release, written as machine-readable decision tables (tables that give an answer for each combination of conditions) for two rule books: the US rule, 21 CFR 610.40, and the UK blood services' testing rules. With them goes a public set of at least 300 synthetic donation cases, each with the correct "release" or "hold" answer and the clause it tests, plus a script that runs any blood-bank program against the set and scores it.
Why it is still open
Both rule books are prose with conditions and exceptions. The US rule requires HIV, hepatitis B and C, HTLV, syphilis, West Nile virus and Chagas testing but lets establishments adopt "an adequate and appropriate alternative testing procedure" for some of them; the UK tests every donation for five infections, tests for HTLV only in new donors and some components, and for malaria after recent travel to a country where it is common. A GitHub search found donor-eligibility apps and a synthetic blood-bank data generator, but no encoded testing rules or release test cases.
Built with what exists
The eCFR's public API for the US text and the JPAC guidelines for the UK; DMN (Decision Model and Notation, a published standard for decision tables), or plain Python tables tested with pytest; the MIT-licensed blood-bank-data-generator on GitHub as a base for synthetic records; an AI model to draft cases clause by clause, each checked by a person. Such a suite could also check AI-written blood-bank code, the way AWS checked its agent-written cloud emulator only by tests (Note: "Code no person read").
First step
Pull the current text of 21 CFR 610.40 from the eCFR API and write the first 20 cases for paragraph (a)(1), the HIV, hepatitis B and hepatitis C requirement.
Done looks like
Every paragraph of 21 CFR 610.40 that bears on release, and every UK rule on the GOV.UK summary, is mapped to at least one case; there are at least 300 cases; two readers who did not write them agree with at least 95% of the expected answers; and scores are published for at least two open-source blood-bank programs.