POLICY
1. The NHS stopped waiting for 4 million women to show up. It sent a kit to their door instead.
The signal: NHS England today began rolling out home HPV self-testing kits to 4 million women in England who are behind on cervical screening. Invitations arrive through the NHS App, text, email, or letter. Women order a swab kit, collect a sample at home, and post it back. If HPV is detected, they're called in for a proper appointment.
The details:
Only 68.8% of eligible women are currently up to date with cervical screening, below the NHS's 80% target. The rollout targets those aged 30 to 65 who've not attended after previous invitations.
The evidence base comes from the YouScreen trial, run by North London cancer alliances with King's College London, which found home self-testing could boost England's screened population by approximately 400,000 per year.
Cervical screening appointments remain the recommended route for those who already attend. This is a parallel offer, not a replacement.
Why it matters: A third of eligible women weren't coming in before today. Mailing them a kit is a simpler intervention than most HealthTech investments and it will almost certainly screen more patients than any clinical AI upgrade rolled out this year. The hard work in cancer prevention has always been logistics, not science.
CLINICAL AI
2. A routine blood test can now generate your full molecular portrait. AI just did the expensive bit for free.
Source: Nature Computational Science. The PULSE framework uses routine clinical data to infer missing multimodal biomarkers across patient visits.
The signal: A team from Macau University published PULSE (Patient Unified Longitudinal Signal Engine) in Nature Computational Science this week. The model reads 61 routine blood markers that any GP orders, and generates 251 full metabolomic biomarkers that normally require expensive specialist laboratory panels. It outperformed all benchmark methods against ground truth UK Biobank data.
The details:
PULSE uses a patient's own longitudinal history as context. The same test result means different things depending on where it is in your individual trajectory: the model learns that pattern rather than treating each visit in isolation.
Beyond metabolomics, it also generated proteomic profiles from standard labs and medical history, with AUC values of 0.72 to 0.83 for disease prediction across six common diseases, matching results from actual proteomic data.
Validated across UK Biobank, US and Chinese ICU cohorts (127,731 patients, 657,103 clinical visits). The framework is open-access and designed as a general foundation rather than a single-task product.
Why it matters: Deep molecular profiling has been limited to well-resourced research centres. If PULSE holds up in prospective validation, every GP practice with standard bloodwork suddenly has access to a layer of biology that currently costs hundreds or thousands per test. That's not a marginal improvement in precision medicine, it changes who can afford to participate in it.