Skip to main content
Back to challenges
Challenge

Postpartum Psychiatric Prior-Authorization Audit Agent with Claude Agents SDK

Reproductive psychiatry treatments involve complex liability and strict authorization rules. Construct a safe clinical reviewer agent using Claude Agents SDK to audit prior-authorization requests against clinical guidelines and risk markers. Achieve 100% audit trail completeness and zero unhandled adverse contraindications.

Data ScienceHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Implement a safe, auditing agent for psychiatric prior-authorization using Claude Agents SDK with strict rule validation.

Delivery guide

How work is evaluated

Evaluation

Evaluates prior-authorization decision accuracy and audit trail compliance.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

audit_completeness

Requires non-empty audit trail array for compliance.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

contraindication_detection_rate

Accuracy in detecting critical clinical drug contraindications • target: 1 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Build extended-thinking clinical safety reviewers using Claude Agents SDK

  • Implement prior-authorization rule checks for high-risk psychiatric drug approvals

  • Verify contraindication flags against maternal health safety databases

  • Enforce deterministic clinical logging for regulatory and liability compliance

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic cohort of 100 postpartum depression and psychosis prior-authorization claims with expert clinical reviewer decisions.

Challenge setup

Evaluates prior-authorization decision accuracy and audit trail compliance.

Challenge input

JSON with patient clinical profile, requested drug, dose, and concurrent medications.

Claude Agents SDK

Framework specified for building safety-focused, high-reasoning agents.

Evaluated output

JSON object with authorization_status (APPROVED|DENIED|REQUIRES_HUMAN_REVIEW), contraindications_flagged, and step_by_step_audit_trail.

Evaluation checks
  • Requires non-empty audit trail array for compliance.
  • Accuracy in detecting critical clinical drug contraindications • target: 1 • range: 0-1
Configured evaluation evidence
  • Benchmark: Prior Authorization Compliance Benchmark
  • Contraindication Detection Rate target: 100%
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Squarespace · Required
  • RAI · Optional
Policy Serving
  • Claude 4 Opus · Optional

Evaluation contract

  • audit_completeness · Weight 1
  • contraindication_detection_rate · Weight 1

Metrics

  • contraindication_detection_rate · Target 1

Benchmarks

  • Prior Authorization Compliance Benchmark
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with patient clinical profile, requested drug, dose, and concurrent medications.

    Output: JSON object with authorization_status (APPROVED|DENIED|REQUIRES_HUMAN_REVIEW), contraindications_flagged, and step_by_step_audit_trail.

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · Agent Architecture Engine

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Postpartum Psychiatric Prior-Authorization Audit Agent with Claude Agents SDK