Case Study: Automating High-Volume Pharma Prior Authorization

Executive Summary

Our client, a leading digital weight-loss clinic, faced a critical operational bottleneck: their rapid growth was outpacing their ability to process medication approvals. With 100,000+ GLP-1 prescriptions annually (e.g., Wegovy, Ozempic), they relied on a team of 50 coordinators spending 20–30 minutes per case manually typing data into payer portals.

We deployed an end-to-end automated Prior Authorization (PA) pipeline that reduced the dedicated team to just 6 coordinators while maintaining a 95% automation rate. The system slashed processing time to under 1 minute for most cases and enabled real-time status synchronization, transforming a multi-day manual slog into an instant background process.

---

The "Before" State: The Manual Bottleneck

Prior to automation, the workflow was labor-intensive and error-prone. A typical case involved 11 manual steps, some of which were:

1. Initiation Delay: A provider approved a medication, creating a task that sat in a queue for days.
2. Triple Data Entry: Coordinators opened three separate windows (EMR, e-Prescribing, PA Portal) and manually copy-pasted patient demographics, insurance BIN/PCN/Group numbers, and clinical notes.
3. Manual Transcription: Staff had to visually read photos of insurance cards to type in policy numbers, leading to frequent typos and "Member Not Found" rejections.
4. Clinical Hunt-and-Peck: Answering payer questionnaires required searching through unstructured clinical notes to find "date of last weigh-in" or "failed diet attempts."
5. Blind Status Checks: Staff had to manually log in to payer portals every 48 hours just to check if a request was approved or denied.

Pain Point: The cost of operations was scaling linearly with patient growth, and delays led to high patient churn.

---

The "After" State: Zero-Touch Automation

We built an event-driven middleware that orchestrates the entire lifecycle of a PA request, from the moment a patient requests medication to the final pharmacy dispatch.

1. Instant Auto-Initiation

Instead of a task queue, the system triggers immediately when a patient requests a brand-name medication and uploads their insurance card.

• Latency: Reduced from days to < 1 second.
• Mechanism: The system creates a "held" (draft) prescription in DoseSpot to generate the necessary medication order object without prematurely transmitting it to a pharmacy. It instantly compiles the PA payload (Patient, Prescriber, Drug, Pharmacy Benefits).

2. Intelligent Data Pre-Population (API-First)

We eliminated manual data entry completely.

• Auto-Complete: The system pulls validated patient demographics and insurance eligibility data automatically (removing the need for visually inspecting insurance cards).
• Document Attachment: Clinical notes and relevant lab results (e.g., HbA1c, Lipid Panel) are automatically generated and attached via API, ensuring 100% of required documentation is present at submission.

3. Smart Question automation

The core of the system is a hybrid Rules Engine and AI Question Answerer.

• Structured Data Rules: Simple questions (e.g., "Is patient's BMI > 30?") are answered instantly using structured vitals data. We implemented granular ICD-10 logic (Z68.x for BMI, E66.x for obesity) to prevent technical denials.
• AI Clinical Justification: For complex free-text questions (e.g., "Describe prior failure on Metformin"), our Medical LLM extracts narrative evidence from the patient's chart, citing specific dates and outcomes to prevent hallucinations.

4. Automated Denial Management & "Waterfall" Logic

A denial triggers an automatic "Plan B" workflow without human intervention. The system follows a strict denial reason-driven playbook. E.g., if 'step therapy required' is returned from a payer, we follow their required "waterfall" based on the patient's condition and payer rules, such as:

• For Weight Loss (Non-T2D):
  1. Submit Wegovy → If Denied:
  2. Auto-create PA for Zepbound → If Denied:
  3. Auto-create PA for Saxenda.
• For Type 2 Diabetes:
  1. Submit Mounjaro → If Denied:
  2. Submit Ozempic → If Denied:
  3. Submit Wegovy.

5. Real-Time Status & Patient Communication

We replaced manual "check-ins" with webhooks that sync status within minutes (down from 48+ hours). These status changes trigger automated patient communication macros:

• On Submission: "We have submitted your paperwork. Expect a response in up to 14 days."
• On Approval: "Great news! Your medication is approved and being sent to the pharmacy."
• On Denial: "Your insurance denied the first option. We are automatically applying for an alternative medication now."

---

Business Impact & Metrics

Metric	Before (Manual)	After (Automated)
Team Size	50 Coordinators	6 Coordinators
Time per Case	20–30 minutes	≤ 3 minutes (human review only)
Field Accuracy	Variable (Typo-prone)	≥ 99% (Source-of-truth sync)
Status Sync	48 hours (Manual)	≤ 4 hours (Real-time Webhook)
Prescription Safety	Rx often written before approval	Rx held in draft until PA approved

By automating the routine 95% of prior authorizations, the client successfully scaled their weight-loss program without exploding their operational costs, ensuring patients receive life-changing medication faster and with less friction.