AI Agent for Public Health: Vaccination Information, Clinic Finder, and Outbreak Alerts

Public Health Information at Scale

Public health departments serve as the front line of community health infrastructure. They manage vaccination programs, track disease outbreaks, operate clinics, and communicate health advisories. During routine operations, residents call with questions about vaccine eligibility, clinic hours, and immunization records. During outbreaks, call volumes spike by 10x or more, overwhelming staff and leaving residents without timely information.

An AI agent can handle the information layer: determining vaccine eligibility based on age and health conditions, finding nearby clinics with availability, checking immunization records, and distributing outbreak alerts with specific guidance. The agent does not replace clinical judgment — it replaces the phone tree and the hold queue.

Vaccine Eligibility Engine

Vaccine eligibility rules are set by the CDC and state health departments. They follow structured criteria based on age, health conditions, and prior vaccination history. This is deterministic logic that must be precise.

flowchart LR
    CALLER(["Patient or Caregiver"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Healthcare AI Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["Appointment booked"])
        O2(["Prescription refill request"])
        O3(["Triage to clinician"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

from dataclasses import dataclass, field
from datetime import date
from enum import Enum

class VaccineType(Enum):
    FLU = "influenza"
    COVID = "covid_19"
    TDAP = "tdap"
    MMR = "mmr"
    SHINGLES = "shingles"
    PNEUMOCOCCAL = "pneumococcal"
    HPV = "hpv"
    HEPATITIS_B = "hepatitis_b"

@dataclass
class VaccineRule:
    vaccine: VaccineType
    min_age: int | None = None
    max_age: int | None = None
    recommended_for: list[str] = field(default_factory=list)
    contraindications: list[str] = field(default_factory=list)
    dose_schedule: str = ""
    requires_prior_dose: bool = False
    seasonal: bool = False

VACCINE_SCHEDULE: dict[VaccineType, VaccineRule] = {
    VaccineType.FLU: VaccineRule(
        vaccine=VaccineType.FLU,
        min_age=6,  # months, but we simplify to years here
        recommended_for=["everyone 6 months and older"],
        contraindications=["severe egg allergy (egg-free options available)"],
        dose_schedule="Annually, typically September through March",
        seasonal=True,
    ),
    VaccineType.SHINGLES: VaccineRule(
        vaccine=VaccineType.SHINGLES,
        min_age=50,
        recommended_for=[
            "adults 50 and older",
            "adults 19+ with weakened immune systems",
        ],
        dose_schedule="Two doses, 2-6 months apart (Shingrix)",
    ),
    VaccineType.HPV: VaccineRule(
        vaccine=VaccineType.HPV,
        min_age=9,
        max_age=45,
        recommended_for=[
            "routine: ages 11-12 (can start at 9)",
            "catch-up: through age 26",
            "shared decision: ages 27-45",
        ],
        dose_schedule="2 doses if started before 15; 3 doses if started at 15+",
    ),
    VaccineType.PNEUMOCOCCAL: VaccineRule(
        vaccine=VaccineType.PNEUMOCOCCAL,
        min_age=65,
        recommended_for=[
            "adults 65 and older",
            "adults 19-64 with certain medical conditions",
            "adults 19-64 who smoke",
        ],
        dose_schedule="PCV20 single dose, or PCV15 followed by PPSV23",
    ),
}

@dataclass
class PersonProfile:
    age: int
    conditions: list[str] = field(default_factory=list)
    prior_vaccines: dict[str, date] = field(default_factory=dict)
    pregnant: bool = False
    immunocompromised: bool = False

def check_vaccine_eligibility(
    person: PersonProfile,
) -> list[dict]:
    """Check which vaccines a person is eligible for."""
    results = []

    for vtype, rule in VACCINE_SCHEDULE.items():
        eligible = True
        reasons = []

        # Age check
        if rule.min_age and person.age < rule.min_age:
            eligible = False
            reasons.append(f"Minimum age is {rule.min_age}")
        if rule.max_age and person.age > rule.max_age:
            eligible = False
            reasons.append(f"Maximum age is {rule.max_age}")

        # Contraindication check
        for contra in rule.contraindications:
            if any(c.lower() in contra.lower() for c in person.conditions):
                reasons.append(f"Contraindication: {contra}")

        # Check if already vaccinated recently
        last_dose = person.prior_vaccines.get(vtype.value)
        if last_dose and not rule.seasonal:
            reasons.append(f"Last dose: {last_dose.isoformat()}")

        results.append({
            "vaccine": vtype.value,
            "eligible": eligible,
            "recommended_for": rule.recommended_for,
            "reasons": reasons,
            "dose_schedule": rule.dose_schedule,
        })

    return results

Clinic Finder with Availability

Finding the right clinic involves geographic proximity, vaccine availability, appointment slots, and sometimes insurance acceptance. The agent queries a clinic database and returns actionable options.

from dataclasses import dataclass, field
from math import radians, sin, cos, sqrt, atan2

@dataclass
class Clinic:
    clinic_id: str
    name: str
    address: str
    latitude: float
    longitude: float
    phone: str
    hours: dict[str, str]  # day -> "9:00 AM - 5:00 PM"
    vaccines_available: list[VaccineType] = field(default_factory=list)
    accepts_walkins: bool = False
    accepts_insurance: list[str] = field(default_factory=list)
    next_available_slot: str | None = None

def haversine_distance(lat1: float, lon1: float, lat2: float, lon2: float) -> float:
    """Calculate distance between two points in miles."""
    R = 3959  # Earth radius in miles
    dlat = radians(lat2 - lat1)
    dlon = radians(lon2 - lon1)
    a = sin(dlat / 2) ** 2 + cos(radians(lat1)) * cos(radians(lat2)) * sin(dlon / 2) ** 2
    return R * 2 * atan2(sqrt(a), sqrt(1 - a))

def find_nearby_clinics(
    user_lat: float,
    user_lon: float,
    vaccine_needed: VaccineType | None = None,
    max_distance_miles: float = 10.0,
    clinics: list[Clinic] = None,
) -> list[dict]:
    """Find clinics near the user, optionally filtered by vaccine availability."""
    results = []

    for clinic in clinics or []:
        distance = haversine_distance(user_lat, user_lon, clinic.latitude, clinic.longitude)
        if distance > max_distance_miles:
            continue

        if vaccine_needed and vaccine_needed not in clinic.vaccines_available:
            continue

        results.append({
            "name": clinic.name,
            "address": clinic.address,
            "phone": clinic.phone,
            "distance_miles": round(distance, 1),
            "walk_ins": clinic.accepts_walkins,
            "next_appointment": clinic.next_available_slot,
            "vaccines": [v.value for v in clinic.vaccines_available],
        })

    results.sort(key=lambda x: x["distance_miles"])
    return results[:10]

Outbreak Alert Distribution

During disease outbreaks, the agent becomes a critical communication channel. It must distribute accurate, actionable information without causing panic.

from datetime import datetime

@dataclass
class OutbreakAlert:
    alert_id: str
    disease: str
    severity: str  # low, moderate, high, critical
    affected_areas: list[str]
    case_count: int
    date_issued: datetime
    summary: str
    prevention_steps: list[str]
    symptoms_to_watch: list[str]
    when_to_seek_care: str
    exposure_locations: list[dict] | None = None

ACTIVE_ALERTS: list[OutbreakAlert] = []

def get_relevant_alerts(user_zipcode: str) -> list[dict]:
    """Get outbreak alerts relevant to the user's location."""
    relevant = []
    for alert in ACTIVE_ALERTS:
        if user_zipcode in alert.affected_areas or "all" in alert.affected_areas:
            relevant.append({
                "disease": alert.disease,
                "severity": alert.severity,
                "case_count": alert.case_count,
                "summary": alert.summary,
                "prevention": alert.prevention_steps,
                "symptoms": alert.symptoms_to_watch,
                "seek_care_when": alert.when_to_seek_care,
                "date_issued": alert.date_issued.isoformat(),
            })

    # Sort critical alerts first
    severity_order = {"critical": 0, "high": 1, "moderate": 2, "low": 3}
    relevant.sort(key=lambda x: severity_order.get(x["severity"], 4))
    return relevant

Agent Prompt Design for Public Health

The public health agent prompt must balance helpfulness with medical accuracy. It provides information from authoritative sources (CDC, state health department) and always directs clinical questions to healthcare providers.

HEALTH_AGENT_PROMPT = """You are a public health information assistant for
the county health department.

You help residents with:
- Vaccine eligibility and scheduling
- Finding nearby clinics
- Understanding outbreak alerts
- Immunization record questions

RULES:
1. Base all vaccine information on CDC recommendations.
2. Never diagnose conditions or recommend treatments.
3. For symptoms: provide general guidance and direct to a healthcare provider.
4. For outbreak alerts: share facts, prevention steps, and when to seek care.
5. Always mention that individual medical decisions should involve a doctor.
6. If someone describes an emergency, direct them to call 911 immediately.
"""

FAQ

How does the agent handle misinformation about vaccines?

The agent responds to misinformation with factual, evidence-based information from the CDC and peer-reviewed sources. It does not argue or become confrontational. For example, if a user says "vaccines cause autism," the agent responds: "Extensive research involving millions of children has found no link between vaccines and autism. The original study claiming this link was retracted due to serious methodological flaws. I can share links to the CDC's vaccine safety research if you would like to review the evidence." The agent acknowledges the person's concern, provides facts, and offers resources.

How does the agent maintain accuracy as vaccine recommendations change?

Vaccine recommendations are stored in a versioned configuration that is updated whenever the CDC issues new guidance. The agent's eligibility engine reads from this configuration at runtime, so updates take effect immediately. A public health administrator reviews and approves all changes before they go live. The agent includes a "last updated" timestamp in eligibility responses so users know how current the information is.

Can the agent help parents track their children's immunization schedules?

Yes. The agent can look up a child's immunization record (with parental consent and identity verification), compare it against the CDC recommended schedule for the child's age, and identify which vaccines are due or overdue. It generates a personalized schedule showing what vaccines are needed at upcoming well-child visits. The agent can also send reminders when the next dose is due, reducing missed vaccinations.

---

#GovernmentAI #PublicHealth #Vaccination #HealthAlerts #ClinicFinder #AgenticAI #LearnAI #AIEngineering