OrthoPath

Assistive Clinical Decision Support
Assistive-only CDS: OrthoPath provides clinical decision support only. Clinicians retain full responsibility for all clinical decisions. This tool does not autonomously diagnose, prescribe, or order. Do not enter patient identifiers or any Protected Health Information (PHI) into this MVP.

About OrthoPath

Last updated: 2026-02-28

Assistive-Only Disclaimer

OrthoPath is an assistive clinical decision support (CDS) tool only. It does not autonomously diagnose, prescribe, order tests, or guide discharge. All outputs require independent clinical review by a qualified clinician. The clinician retains full professional and legal responsibility for all clinical decisions.

Do not enter patient identifiers or any Protected Health Information (PHI) into this tool.

Authorship and Attribution

All clinical logic was authored, reviewed, and validated by Dr. Pendar. Evidence anchors for each module are drawn from published guidelines and peer-reviewed literature cited in the per-module evidence manifests. No rule fires without a traceable evidence source.

This system has not undergone regulatory clearance (e.g. FDA 510(k), CE marking). It is provided as an assistive reference tool only.

What OrthoPath Is

OrthoPath is a deterministic, rules-based assistive CDS system for selected orthopaedic clinical questions. It evaluates structured clinical inputs against evidence-anchored rule sets and returns a structured suggestion with the rules that fired and the evidence sources supporting them. The system converts curated evidence and pathway logic into computer-readable rules. Its purpose is to help clinicians apply orthopaedic protocols consistently by providing structured outputs and clear rationale—while remaining clinician-in-the-loop.

Intended Users and Setting

  • Intended users: Qualified orthopaedic surgeons, emergency physicians, anaesthesiologists, and other licensed clinicians.
  • Setting: Secondary or tertiary care, point-of-care use alongside standard clinical assessment.
  • Not for: Autonomous use by non-clinicians. Not a substitute for specialist consultation.

System Architecture

OrthoPath is a monorepo comprising three services:

  • API Gateway — Tenant-aware HTTP API (FastAPI): request validation, routing, run persistence, and protocol listing. No recommendation logic; forwards CDS requests to the engine.
  • CDS Engine — Deterministic rule evaluator: loads YAML protocol packs, evaluates each rule’s `when` condition against request inputs, and is the sole service that constructs recommendation objects. Supports priority-ordered firing (P1–P6) and exclusivity-group merge for multi-hit strategy packs.
  • Evidence Service — Read-only retrieval and citation packaging; structurally prevented from emitting recommendations. Used for evidence display and rationale lookup.

Clinical rules are encoded in YAML protocol packs. Each rule’s `when` condition is expressed in CEL (Common Expression Language) or a CEL-compatible subset—non-Turing-complete, safe for embedding—and each pack ships with an evidence manifest linking every rule to its supporting literature.

Limitations

  • Module scope: Six clinical modules are available: Ottawa Ankle & Knee Imaging Rules, Regional Anesthesia & Anticoagulation, Open Fracture Initial Management & Antibiotics, Pathologic Fracture Risk & Treatment (Mirels), Metastatic Bone Disease (MBD) Management — Pro (long bones + reassessment), and ETC vs DCO — Physiology-Guided Fixation Timing (Core snapshot). Module 5 (MBD Pro) is limited to femur and humerus; spine, pelvis, tibia, and other sites receive escalation/MDT referral only. Module 6 (ETC vs DCO) provides assistive fixation-timing support (P1 hard-stops, P3 missingness, P4 ETC permissible, P5 escalation) from a single-encounter physiology snapshot; it does not use timeline or trend data.
  • Evidence-based rules: Rules are based on the evidence cited in each module's evidence manifest. They do not cover all clinical presentations. Humerus Mirels thresholds (e.g. ≥8 vs 7) are evidence-dependent and surfaced as assistive options; MDT discussion is encouraged.
  • Derived features (Module 5): Timeline-derived values (e.g. followup_overdue_flag, pain_trend_over_14d) are v0.1 approximations from structured events and fixed windows; they do not replace a scheduling engine or full clinical documentation.
  • No ML or adaptation: The system does not learn or adapt from clinical use (no machine-learning inference).
  • No EHR integration: The system does not have access to electronic health records (EHR) or real-time patient data. Do not enter PHI; all inputs are structured (no free-text clinical notes as engine input).
  • Assistive only: Outputs are suggestions for consideration. All outputs require independent clinical review in the context of the full clinical picture. The system does not order tests, prescribe, or diagnose.