Appearance
Healthcare Use Cases
Real-world scenarios demonstrating how GhostAI protects PHI while enabling AI-assisted workflows.
Clinical Documentation
Drafting Clinical Notes
Scenario: A physician needs help drafting a discharge summary.
Without GhostAI: PHI would be sent directly to the AI, creating HIPAA liability.
With GhostAI:
User Input:
Help me write a discharge summary for:
Patient: John Smith, MRN: 12345678
DOB: 01/15/1960, Admitted: 03/15/2024
Diagnosis: Acute MI, underwent PCI to LAD
Complications: None
Follow-up: Cardiology in 2 weeksWhat AI Sees:
Help me write a discharge summary for:
Patient: [PERSON_a1b2], MRN: [MRN_c3d4]
DOB: [DATE_e5f6], Admitted: [DATE_g7h8]
Diagnosis: Acute MI, underwent PCI to LAD
Complications: None
Follow-up: Cardiology in 2 weeksAI Response (detokenized):
DISCHARGE SUMMARY
Patient: John Smith
MRN: 12345678
DOB: 01/15/1960
Admission Date: 03/15/2024
DIAGNOSIS: Acute Myocardial Infarction
HOSPITAL COURSE:
John Smith was admitted with acute MI and successfully
underwent percutaneous coronary intervention to the LAD.
Recovery was uncomplicated.
FOLLOW-UP:
- Cardiology appointment in 2 weeks
- Continue dual antiplatelet therapy
- Cardiac rehabilitation referralProgress Notes
Scenario: Nurse documenting daily progress.
Input:
Write a progress note for Mary Johnson, room 302, day 2 post-op.
Vitals stable, pain 3/10, ambulating with assistance.
Tolerating clear liquids.Protected Output: AI generates proper progress note format with Mary Johnson's name restored, while never seeing the actual PHI during processing.
Patient Communication
Appointment Reminders
Scenario: Staff drafting personalized appointment reminders.
Input:
Write a friendly reminder email for:
Patient: Robert Brown
Appointment: March 25, 2024 at 2:30 PM
Provider: Dr. Sarah Wilson
Location: Main Campus, Suite 200Result: AI drafts email with protected PHI, then GhostAI restores patient name for the final output.
Patient Education
Scenario: Creating customized patient education materials.
Input:
Create diabetes education materials for Jane Doe,
newly diagnosed with Type 2 diabetes.
A1C: 8.2%, currently starting on Metformin.Result: Personalized education materials without exposing patient data to AI.
Medical Coding & Billing
ICD-10 Code Assistance
Scenario: Coder needs help with complex diagnosis coding.
Input:
What ICD-10 codes apply for a patient with:
- CHF with preserved ejection fraction
- Type 2 diabetes with CKD stage 3
- HypertensionNote: No PHI in this query, so it passes through normally. GhostAI is smart enough not to tokenize clinical terminology.
Denial Management
Scenario: Appealing insurance denials.
Input:
Draft an appeal letter for claim denial:
Patient: Michael Chen, DOB: 05/20/1975
Claim #: 123456789
Service: MRI Lumbar Spine
Denial reason: Not medically necessary
Clinical notes: Chronic low back pain x 6 months, failed PTResult: Professional appeal letter with patient identifiers protected during AI processing.
Care Coordination
Referral Letters
Scenario: PCP referring patient to specialist.
Input:
Write a referral to cardiology for:
Patient: Susan Williams, 67F
Chief concern: Exertional dyspnea
History: HTN, HLD, former smoker
Recent findings: Abnormal stress testOutput: Properly formatted referral letter with all PHI restored.
Care Transition Summaries
Scenario: Transferring patient to skilled nursing facility.
Input:
Create a care transition summary for:
Patient: George Thompson
Transferring to: Sunny Meadows SNF
Current meds: [list]
Care needs: Wound care, PT/OT, medication management
Contact: Daughter - (555) 123-4567Result: Comprehensive transfer summary with protected family contact info.
Quality & Research
Case Reviews
Scenario: Peer review of clinical outcomes.
Input:
Help me analyze this case for M&M conference:
Patient: 72yo male, admitted with sepsis
Initial presentation: [details]
Outcome: [details]
What could have been done differently?Protection: Patient identifiers are tokenized; clinical details remain for educational discussion.
Quality Metrics
Scenario: Analyzing readmission patterns.
Input:
Summarize factors contributing to 30-day readmission for:
Patient: James Wilson
Original admission: CHF exacerbation
Readmission: 21 days later with same
Social factors: Lives alone, limited supportResult: Quality analysis with patient privacy preserved.
Administrative Tasks
Prior Authorization
Scenario: Drafting prior auth requests.
Input:
Write prior auth request for:
Patient: Lisa Garcia, DOB: 08/30/1985
Insurance: BlueCross ID: XYZ123456
Requested: Humira for moderate RA
Clinical justification: Failed MTX and PlaquenilOutput: Insurance-ready prior auth with identifiers protected.
Insurance Verification Letters
Scenario: Verifying coverage for procedures.
Input:
Draft letter confirming medical necessity for:
Patient: David Kim
Procedure: Bariatric surgery
BMI: 42, comorbidities: T2DM, HTN, OSA
Previous attempts: Diet, exercise, medicationsResult: Professional letter supporting medical necessity.
Emergency Scenarios
Rapid Assessment
Scenario: ED physician needs quick differential diagnosis help.
Input:
65yo female presents with sudden severe headache,
photophobia, neck stiffness. BP 180/100.
What should I consider?Note: No PHI here - clinical scenario without identifiers. GhostAI allows this through without tokenization.
Transfer Coordination
Scenario: Coordinating critical care transfer.
Input:
Prepare transfer summary for:
Patient: 45yo male, MRN 98765432
Diagnosis: STEMI
Transferring to: Regional Medical Center
Contact physician: Dr. Jones at (555) 234-5678Result: Complete transfer documentation with protected identifiers.
Telehealth Support
Visit Summaries
Scenario: Documenting telehealth visits.
Input:
Summarize telehealth visit for:
Patient: Patricia Lee
Visit type: Follow-up diabetes management
Discussed: Medication adjustment, lifestyle
Next steps: Labs in 2 weeks, f/u in 1 monthOutput: Professional visit summary for patient portal.
After Visit Instructions
Scenario: Creating personalized discharge instructions.
Input:
Create after-visit instructions for:
Patient: Thomas Wright
Diagnosis: Acute bronchitis
Medications: Z-pack, Tessalon
Follow-up: Return if symptoms worsenResult: Clear, personalized instructions with protected PHI.
Best Practices Summary
| Use Case | PHI Involved | Protection Level |
|---|---|---|
| Clinical notes | High | Full tokenization |
| Patient communication | Medium | Name + contact |
| Coding queries | None | Pass through |
| Care coordination | High | Full tokenization |
| Quality analysis | Medium | Identifiers only |
| Administrative | High | Full tokenization |
Getting Started
- Set vertical: Settings → Healthcare
- Enable PHI detection: Verify all 18 identifiers
- Review thresholds: Adjust for your workflow
- Train staff: Share this documentation
- Monitor compliance: Review audit logs weekly
Prompt Enhancement
The magic wand automatically frames prompts with HIPAA awareness and clinical terminology.
Example: Clinical Interpretation
You type:
explain this lab resultEnhanced:
As a clinical professional, I need you to interpret the following lab result.
Explain each biomarker's significance, flag any values outside normal ranges,
suggest potential differential diagnoses, and recommend follow-up tests.
Present findings in a structured clinical note format suitable for the patient chart.PII Protection: Patient identifiers in the lab result are tokenized before the enhanced prompt is sent to the AI.