Skip to content

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 weeks

What 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 weeks

AI 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 referral

Progress 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 200

Result: 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
- Hypertension

Note: 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 PT

Result: 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 test

Output: 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) 123-4567

Result: 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 support

Result: 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 Plaquenil

Output: 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, medications

Result: 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) 234-5678

Result: 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 month

Output: 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 worsen

Result: Clear, personalized instructions with protected PHI.

Best Practices Summary ​

Use CasePHI InvolvedProtection Level
Clinical notesHighFull tokenization
Patient communicationMediumName + contact
Coding queriesNonePass through
Care coordinationHighFull tokenization
Quality analysisMediumIdentifiers only
AdministrativeHighFull tokenization

Getting Started ​

  1. Set vertical: Settings → Healthcare
  2. Enable PHI detection: Verify all 18 identifiers
  3. Review thresholds: Adjust for your workflow
  4. Train staff: Share this documentation
  5. 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 result

Enhanced:

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.

SOC 2 Type II Certified | BAA Available | HIPAA-Ready