Surveillance Use Cases

AI Watching the Literature

How do you build an AI-assisted evidence surveillance system that keeps a WHO India brief current — without running a full review every time something changes?

The evidence half-life problem

A well-conducted literature review has a shelf life. For health financing and UHC topics in India, that shelf life is often shorter than the policy cycle it is meant to inform. PM-JAY implementation data accumulates quarterly. NHA accounts are released annually. HTAIn publishes new assessments on a rolling basis. Budget Union statements shift the fiscal landscape. A brief produced in January may be materially outdated by the time it reaches a decision-maker in August.

The conventional response — running a new full systematic review whenever the evidence base shifts — is not realistic at WHO India's operating tempo. Evidence surveillance is the alternative: a structured, lightweight monitoring system that detects when new evidence is significant enough to update a brief, without requiring a full review to find out.

AI tools are well-suited to surveillance. They can monitor keyword combinations, flag new publications, screen abstracts against your inclusion criteria, and summarise what has changed. The challenge is not the technology — it is designing the surveillance protocol so that the signal-to-noise ratio stays high enough to be actionable rather than overwhelming.

Three tiers of surveillance intensity

Not every topic warrants the same surveillance investment. Calibrate by how fast the evidence base moves and how consequential a missed update would be:

Tier 1

Passive monitoring

Monthly · 15 minutes

Set-and-forget Google Scholar alerts and Semantic Scholar email notifications for your key search terms. Review the digest once a month. Flag anything that meets your PECO-F criteria for closer reading.

Tools: Google Scholar alerts · Semantic Scholar · Research Rabbit updates

Tier 2

Active triage

Fortnightly · 45 minutes

Run your saved search strings in PubMed and Dimensions.ai with a date filter for the last two weeks. Paste the abstract list into Claude or Elicit with a screening prompt. Review only the flagged abstracts yourself. Supplement with manual checks of HTAIn, MoHFW, and NITI Aayog for new grey literature.

Tools: PubMed saved searches · Elicit · Claude · HTAIn manual browse

Tier 3

Living evidence tracking

Weekly · 90 minutes

Reserved for active policy decisions with a tight timeline — a financing reform under deliberation, a budget submission window, an active HTAIn assessment. Run the full protocol at weekly cadence. Every new finding is immediately assessed against the current brief's conclusions for material impact.

Tools: Full protocol stack · NotebookLM for document comparison · Claude for impact assessment

What to watch — the India-specific signal list

Generic keyword alerts produce noise. For WHO India health financing work, the high-value signals are specific. Monitor these sources and triggers with the highest priority:

Source	What to watch for	Priority	Update trigger
HTAIn (htain.mohfw.gov.in)	New assessments on any intervention in your brief's topic area	High	Any new HTAIn report overrides LMIC-extrapolated cost-effectiveness evidence
National Health Accounts India	Annual NHA release — OOP expenditure share, state-level financing data	High	Updated OOP figures may change the baseline financial risk picture materially
PM-JAY / NHA (National Health Authority)	Quarterly claims data, state empanelment reports, beneficiary utilisation	High	Claims settlement rates below 60% in a key state change the implementation feasibility evidence
PubMed — India health financing	New economic evaluations, PM-JAY impact studies, state insurance evaluations	Medium	New RCT or well-designed DiD study with India-specific financial protection outcomes
Union Budget / State Budgets	Health allocation changes, NHM funding, PM-JAY budget revision	High	Material change in fiscal space for the intervention under review
WHO IRIS / World Bank OKR	New LMIC systematic reviews on health financing, UHC progress reports	Medium	New systematic review that includes India or South Asia with disaggregated results
NITI Aayog / MoHFW reports	State health index, NHP progress reports, disease burden updates	Routine	New data that changes the population-level burden estimate underlying the brief's framing

From signal to action — a worked example

Surveillance is only useful if it generates a clear decision: does this new finding change the brief, and if so, how? Here is how to process a new signal:

📡 Worked signal — new HTAIn assessment detected

Signal detected

HTAIn publishes a new cost-effectiveness assessment of PM-JAY outpatient coverage expansion for rural BPL households — directly relevant to the brief you produced in Session 5.

Immediate check

Upload the HTAIn PDF to Claude. Ask: "Does this assessment change any of the five findings or the recommendation in [paste brief text]? If yes, which finding changes and in what direction?"

Three outcomes

No material change: findings consistent with brief → log the new source in the protocol, no brief update required.
Strengthens existing finding: brief finding now has India-specific HTAIn support → update the finding's quality caveat from "moderate, extrapolated from LMIC" to "moderate, India HTAIn evidence."
Contradicts brief finding: HTAIn shows no financial protection effect where brief claimed moderate benefit → brief must be updated or withdrawn before next circulation.

Action rule: if a new HTAIn or peer-reviewed India study materially contradicts a brief recommendation, the brief must carry a visible update note before any further circulation. The surveillance system's value is precisely that it catches this before a contradicted recommendation reaches a decision-maker.

Build your surveillance plan

👁️ Surveillance plan builder

Describe a live WHO India topic you are monitoring or a brief you have recently produced. The tool will generate a structured surveillance protocol — sources, cadence, screening prompts, and update triggers — calibrated to your specific topic.

Topic / brief title What evidence base are you monitoring?

Surveillance tier How fast is this evidence base moving?

Key findings to protect The brief conclusions most at risk of being overturned

Known evidence gaps What new evidence would most change the brief?

👁️ Surveillance protocol — ready to implement

✅ Level 2 complete

You can now read AI-retrieved evidence critically.
Level 3 is where it becomes systemic.

You have worked through the five Exploration skills: detecting structural bias in AI outputs, running iterative prompt sequences, appraising economic studies for India usability, judging when statistical pooling is legitimate, and building surveillance systems that keep evidence current. Level 3 — Integration — asks the harder question: how do you embed all of this into institutional practice, navigate equity and ethics, and use AI to support decisions under genuine uncertainty and real political constraints?

Begin Level 3 — Integration →

🎯 Level 2 takeaway

Evidence surveillance is the system that prevents a well-built brief from becoming a liability when the evidence base moves. The three tiers match surveillance intensity to topic velocity. The India-specific signal list prioritises HTAIn, NHA, and PM-JAY operational data above generic PubMed alerts — because those sources move faster and matter more for WHO India policy than any journal publication. Build the surveillance plan before you circulate the brief. The moment a brief reaches a decision-maker is the moment it starts ageing.