Workflow Library

APHI organizes public health AI around defined questions, not generic prompts. Each workflow states what AI can assist, what remains human-led, and how the output should be evaluated.

Surveillance Draft

Signal Review Memo

Question: Is something unusual happening, and what deserves epidemiologist review?

  • Input: surveillance summary, location, time window, baseline, known data quality issues
  • AI assists: summarize the signal, structure uncertainty, draft reviewer questions
  • Human led: confirm case definition, interpret signal, decide escalation
  • Evaluation: false signal review, missed signal review, reviewer acceptance
Outbreak Response Draft

Investigation Triage Packet

Question: What should the response team do next?

  • Input: early case information, exposure hypotheses, lab updates, jurisdiction context
  • AI assists: organize action items, evidence gaps, situation update draft
  • Human led: define official case status, control measures, public guidance
  • Evaluation: completeness, source traceability, escalation accuracy
Evidence Review Draft

Evidence Brief

Question: What does the evidence say, and how reliable is it?

  • Input: public guidance, peer-reviewed studies, preprints, policy documents
  • AI assists: source triage, evidence table draft, uncertainty summary
  • Human led: evidence grading, recommendation approval, conflict resolution
  • Evaluation: citation accuracy, source freshness, unsupported claim checks
Communication Draft

Public Advisory Draft

Question: What needs to be communicated, to whom, and with what caveats?

  • Input: approved facts, audience, local guidance, uncertainty boundaries
  • AI assists: plain-language draft, FAQ variants, consistency checks
  • Human led: approve public guidance, legal review, media strategy
  • Evaluation: clarity, stigma review, factual consistency, readability
Policy Concept

Decision Brief

Question: What options exist, and what evidence supports each option?

  • Input: policy question, evidence base, constraints, implementation context
  • AI assists: option table, tradeoff summary, leadership questions
  • Human led: policy judgment, legal review, final recommendation
  • Evaluation: claim support, uncertainty clarity, option completeness
Program Evaluation Concept

Evaluation Question Builder

Question: Is the program working, for whom, and under what conditions?

  • Input: program goals, intervention logic, indicators, comparison options
  • AI assists: logic model draft, indicator dictionary, qualitative themes
  • Human led: design selection, causal interpretation, program changes
  • Evaluation: utility, causal restraint, subgroup and missingness review

Have a workflow that should be in APHI?

Useful contributions start with a real public health question, a reviewer, and a way to evaluate the output.

Propose a Workflow Review Evaluation Standards