AI Integration for IDEXX Neo Prescription Refills

The refill workflow in IDEXX Neo typically involves a client request, a veterinarian review of the patient's medical record, and a final approval or denial. AI integrates at three key points: First, it can automatically triage incoming refill requests via the Neo API or a connected queue, prioritizing urgent medications (e.g., insulin, heart medication) and flagging requests that fall outside established protocols. Second, an AI agent can be triggered to review the patient's longitudinal history within Neo—checking last exam date, lab results, and previous prescription notes—and draft a summary for the veterinarian. Third, AI can automate client communication, sending status updates or requests for additional information via Neo's messaging channels.

Implementation connects to Neo's Patient, Prescription, and Communication APIs. A typical architecture uses a middleware layer to listen for new RefillRequest objects. An AI service then retrieves the associated PatientRecord, analyzes structured data and unstructured clinical notes, and generates a recommendation (Approve, Deny, Needs Review). This payload, along with the supporting evidence summary, is posted back to a custom object or dashboard within Neo, creating a prioritized approval queue for the veterinarian. For approved refills, the system can automatically generate the client message and log the interaction, maintaining a full audit trail.

Rollout should start with a pilot on a single, high-volume medication type (e.g., flea/tick preventatives) to validate accuracy and clinician trust. Governance is critical: all AI-generated summaries and recommendations must be reviewed and signed off by a licensed DVM before any action is taken in Neo. The system should be designed for continuous feedback, allowing vets to correct recommendations, which retrains the underlying model. This creates a closed-loop system that improves over time while keeping the veterinarian firmly in control of the final medical decision.

The integration connects to IDEXX Neo's Patient Record API and Prescription/Medication modules to access the structured data needed for review. For each refill request, the system pulls the patient's full history, including past prescriptions, lab results, visit notes, and chronic condition flags. This data is formatted into a context payload and sent to a secure inference endpoint, where a configured LLM (e.g., GPT-4, Claude 3) evaluates the request against clinical guidelines and practice-specific rules.

The AI agent returns a structured recommendation—APPROVE, DENY, or FLAG FOR REVIEW—along with a draft client message and internal notes for the veterinarian. Approved requests trigger an automated workflow via IDEXX Neo's Automation Engine or a custom middleware layer to generate the refill authorization, log the activity, and queue a personalized client notification. Flagged requests are routed to a prioritized approval queue within the Neo interface, surfaced with the AI's reasoning and relevant patient data to expedite the DVM's decision.

Governance is built into the data flow. All AI interactions are logged with the original request payload, model reasoning, and final action in an immutable audit trail. The system supports a human-in-the-loop escalation path, where any recommendation can be overridden, and these overrides are fed back as reinforcement learning data to improve future accuracy. Rollout typically begins in a pilot mode, where the AI acts as a copilot, suggesting actions for vet confirmation, before progressing to full automation for low-risk, routine refills.

A production integration for IDEXX Neo refills is built on a vet-in-the-loop architecture. The AI acts as a pre-screening assistant, never as an autonomous authorizer. It analyzes the refill request against the patient's history in Neo—checking last exam date, chronic condition status, and prescription validity—and generates a draft recommendation (Approve, Deny, or Flag for Review). This draft, along with the supporting data points, is placed into a dedicated Veterinarian Approval Queue within Neo, tagged with a confidence score. The final authorization action always requires a veterinarian's login and explicit approval within the native Neo interface, creating a full audit trail.

Security is managed through Neo's existing RBAC (Role-Based Access Control). The integration service uses a dedicated service account with scoped API permissions, typically limited to read access on patient, prescription, and visit records, and write access only to create notes or queue items in the approval workflow. All PHI remains within Neo's environment; the AI service processes de-identified data payloads or uses a secure, ephemeral context window. Every AI-suggested action and subsequent human decision is logged to a separate audit system for compliance reporting and model performance tracking.

A phased rollout is critical for user adoption and risk management. Phase 1 (Pilot) targets a single, high-volume chronic medication (e.g., thyroid or allergy meds) and a small group of veterinarians. The goal is to validate the AI's accuracy in a low-risk setting and refine the approval queue interface. Phase 2 (Expansion) extends to all chronic medications and the full veterinary team, using the learnings to tune confidence thresholds. Phase 3 (Scale) incorporates acute medication refills and links the system to client communication tools for automated status updates. This stepwise approach builds trust, isolates potential issues, and demonstrates tangible time savings—reducing manual chart review from minutes to seconds—before broadening the scope.