Skip to content

Proposed Algorithm Changes (Pregnancy Configuration)

Audience: algorithm authors and clinical collaborators. This summarizes proposed behavior changes based on the pregnancy protocol and current device plan (OmniPod DASH + Dexcom G7). It is intentionally non‑technical.

Summary

The pregnancy configuration keeps the core algorithm structure intact, but exposes clinician-controlled settings for target, meal upfront percentage, and TMAX. These changes aim to improve pregnancy-specific time-in-range while maintaining safety through runtime guardrails, alerts, and fallback behavior.

Proposed Changes vs. Standard Configuration

1) Clinician-configurable glucose targets - Expose target selection at 90, 100, 110, 120, 130 mg/dL. - Selection step size is 10 mg/dL.

2) Clinician-configurable meal upfront percentage - Expose meal upfront options: - 75% (lower upfront delivery) - 90% (more aggressive upfront delivery) - Rationale: support protocol/clinical adjustment without changing core algorithm structure.

3) Clinician-configurable TMAX - Expose TMAX selection from 40 to 70 minutes, in increments of 5 minutes. - Team-selected default remains configurable in app settings.

4) Unchanged core algorithm structure - No new controllers or major structural changes are proposed. - The algorithm still adapts continuously based on observed glucose and insulin use.

5) CGM downtime behavior retained - If CGM is unavailable, allow fingerstick BG as CGM input. - Basal fallback uses weight‑based dosing early, then adaptive basal after sufficient history (>= 24 hours).

6) Settings governance - These controls are intended to live in a clinician-gated settings area (passcode-gated in current development phase). - Subject ID and weight are part of the same clinician-controlled configuration area.

Operational Context (Non‑Algorithm)

These are delivery‑system behaviors to support safe operation but are not algorithm changes: - Offline mode: if no valid CGM + algorithm run for ≥ 15 minutes, fall back to a predefined basal on the pump and notify the user. - When CGM resumes, algorithm outputs immediately override offline basal. - Device integration is via Dexcom G7 and OmniPod DASH (no Beta Bionics hardware).

Open Questions for Review

  • Should one default target/upfront/TMAX profile be applied at session start, or should users be required to confirm clinician settings before arming?
  • Are additional safety limits needed around when target/upfront/TMAX changes can be applied (for example, only between steps)?
  • Are additional safeguards needed during rapid insulin-resistance changes in late pregnancy?

Status

These are proposed changes derived from the pregnancy protocol and have not been finalized or validated in clinical use within this app.

Pregnancy-configuration algorithm verification is tracked in: - Docs/Planning/Algo2015VerificationPlan.md - Docs/Quality/SoftwareVerificationAndValidationPlan.md (TV-ALG-009 in combination with TV-ALG-001..010) - Docs/Quality/TraceabilityMatrix.md (RA-014)