Pediatrics Clinical Feedback Template

A non-provisioned evaluation template that configures the AI feedback pipeline for pediatric clinical transcripts. Uses a 6-pass analysis architecture grounded in ACGME Pediatrics Milestones, AAP Bright Futures guidelines, and PALLIQS frameworks.

Analysis Passes

Pass	Weight	What it evaluates
Pediatric Assessment	25%	Age-appropriate history taking, developmental screening (ASQ-3/PEDS), growth parameters (WHO/CDC), immunizations, anticipatory guidance (Bright Futures), physical exam technique
Developmental & Behavioral	20%	Developmental surveillance, autism screening (M-CHAT-R), behavioral health (PHQ-A/PSC), school performance, social determinants, ACEs awareness, age-appropriate counseling
Family-Centered Care	20%	Triadic communication, age-appropriate engagement, adolescent confidentiality, parental concern validation, cultural sensitivity, shared decision-making, health literacy
Clinical Reasoning	15%	Age-stratified differentials, weight-based dosing, growth velocity interpretation, vaccination counseling, AAP guideline application, referral decisions
Safety & Prevention	15%*	Medication safety, injury prevention, safe sleep, child welfare screening, HEEADSSS adolescent screening, sport/concussion safety
Synthesis	---	Overall score, EPA level, strengths, improvements, learning goals, supervision recommendations

*Safety & prevention can override the overall rating if critical issues are found (e.g., dosing errors, missed child welfare concerns).

Score Thresholds

Score	Level	Description
9.0+	Distinguished	Teaching/leadership level
7.5+	Proficient	Independent practice ready
6.0+	Competent	Meets expectations for level
4.0+	Developing	Needs additional training
0-4.0	Foundational	Requires close supervision

Setup

export LATTICE_URL=https://access.latticeruntime.com
export LATTICE_TOKEN=$(lattice tokens create)
./scripts/setup-evaluation-template.sh

Configuration

All parameters are mutable and can be overridden per evaluation run:

• provider: anthropic or openai
• model: e.g. claude-sonnet-4-5-20250929, gpt-4o
• temperature: 0.0-1.0 (default: 0.25 for consistent scoring)
• system_prompt: AI evaluator identity and pediatric expertise
• context_prompt: Clinical setting and learner context
• analysis_passes: JSON array defining the 6-pass scoring pipeline
• score_thresholds: Performance level classification rules
• validation_rules: Cross-pass consistency checks

Requirements

The server must have an AI provider API key set:

• ANTHROPIC_API_KEY for Claude models
• OPENAI_API_KEY for GPT models