Care Plan Generation System

Documentation of the four-form assessment system and AI-powered care plan generation process.

System Overview

4

Assessment Forms

Data collection across medical, functional, environmental, and person-centered domains.

AI Processing

Structured prompt combines all data for comprehensive analysis with pattern recognition.

10

Output Documents

Tailored deliverables for caregivers, families, clinical partners, and internal teams.

Process Time: Traditional care plans require 2-4 hours of manual synthesis. This system generates 10 documents in approximately 5 minutes by structuring assessments to feed directly into AI analysis.

Assessment Forms

Four forms capture different dimensions of client needs to create a complete assessment.

Intake Step 1: Phone Intake

Staff-completed form containing PHI

STAFF USE ONLY

Data Collected

  • • Client demographics (name, DOB, age, gender, address, phone, email, SSN last 4)
  • • Payment source and insurance information
  • • Emergency and backup contacts
  • • Primary contact and authorization
  • • Power of attorney information
  • • Medical providers
  • • Home access information

Purpose

Completed by staff during initial phone contact. Contains all PHI required for agency operations, billing, and emergency response. Generates de-identified client code used in all subsequent forms.

HIPAA Note: This is the only form in the system that contains PHI. Client code generated here is used to link all other assessment forms without exposing identifying information.

Intake Step 2: Client Profile

De-identified demographic and medical baseline

Data Collected

  • • Client code (assigned from Step 1)
  • • Care needs overview (free-text + checkboxes)
  • • Schedule requirements (days, shift times, start date)
  • • Living environment (pets, smoking, access methods, safety factors)
  • • Clinical services context (skilled nursing, hospice, wound care, oxygen)

Purpose

Establishes baseline care context, scheduling parameters, and home environment. Links to PHI via client code without containing identifying information.

Format: JotForm with demo button for testing

Person-Centered Survey

Understanding the person beyond diagnosis

Data Collected

  • • Medical conditions and their daily impact
  • • Medications (count, organization, compliance challenges)
  • • Living situation and time spent alone
  • • Support network and family involvement
  • • Recent life changes (losses, transitions, role changes)
  • • Mental health history and current support
  • • Personal background (story, important people, what matters)
  • • Daily life (routines, enjoyments, good vs. difficult days)
  • • Strengths and goals

Purpose

Captures qualitative context for AI pattern recognition. Identifies grief, isolation, medication non-compliance triggers, and behavioral patterns. Informs communication approach and reveals hidden risks.

Format: Optional sections, all fields can be skipped, adapts based on indicated concerns

Functional Assessment

In-person care evaluation using task code scripts

Assessment Protocol

Assessor follows exact task code scripts, asking one question at a time. For each task: document what happens without help, days/times needed, and level of care (0-4 scale).

Level of Care Scale: 0 = No Help, 1 = Reminders, 2 = Supervision, 3 = Standby Assistance, 4 = Hands-On Assistance

24 Task Categories: Meal prep, housework, finances, medications, shopping, transportation, hygiene, dressing (upper/lower), locomotion, transfers, toilet use, bed mobility, eating, bladder/bowel incontinence, personal care, bathing (wash-up/shower), lotion, laundry, reading/writing, supervision

Purpose

Generates shift-specific task lists based on Level of Care ratings and scheduling requirements. Identifies fall risks, transfer safety needs, incontinence management, and supervision requirements. Determines caregiver training needs and equipment requirements.

Format: Assessor-administered with structured interview protocol

Environmental Assessment

Home safety evaluation during in-person visit

Areas Evaluated

  • • Home exterior and entry (steps, handrails, lighting, thresholds)
  • • General interior (cleanliness, trip hazards, clutter, lighting, smoke detectors)
  • • Bathroom - high risk area (tub type, grab bars, surfaces, toilet height)
  • • Bedroom (bed height, path to bathroom, nightlights, medical equipment)
  • • Kitchen (stove safety, accessibility, expired food, fire hazards)
  • • Medications and medical equipment storage
  • • Clinical observations (mobility, cognition, mood, hygiene, physical concerns)

Purpose

Generates equipment recommendations (grab bars, shower bench, raised toilet seat, etc.). Informs caregivers of hazards and home-specific safety protocols. Assigns overall home safety risk level (Low/Moderate/High) to influence supervision level and fall prevention strategies.

Format: On-site inspection focusing on path of travel: bedroom → bathroom → kitchen → exit

AI Prompt Structure

How the four assessment forms combine into a structured prompt for AI analysis.

1. Data Collection

Each form has a "Generate AI Summary" button that creates de-identified export with assessment data only.

2. Prompt Assembly

All four exports combine automatically into a single structured prompt. Demo buttons allow testing with pre-filled data.

3. AI Generation

Prompt instructs AI to analyze patterns, identify risks, and generate 10 tailored output documents.

Prompt Components

Context & Instructions

Establishes AI role as care planning specialist for non-medical home care serving Medicaid clients in Pennsylvania.

Input Data (4 Sections)

Clearly labeled sections for each assessment form's output.

Client Profile Data

Demographics, care needs, schedule, environment

Person-Centered Data

Medical history, preferences, goals, life story

Functional Assessment Data

24 task categories with level of care ratings

Environmental Data

Home safety evaluation and risk assessment

Analysis Framework (10 Components)

AI analyzes data across 10 specific dimensions before generating outputs.

Risk Dashboard
Pattern Analysis
Hidden Risks
Strengths & Goals
Shift Checklists
Escalation Tree
Communication Guide
Care Plan Narrative
Watch List & Meds
Equipment & Environment

Output Format (5 Documents)

Specifies exact formatting and content requirements for each audience-specific document.

  • Caregiver Guide: Shift-by-shift checklists, communication scripts, watch list
  • Client Welcome Packet: Introduction to caregivers and what to expect
  • Family Summary: Overview of care approach without clinical detail
  • Admin Overview: Scheduling, compliance, staffing requirements
  • Clinical Summary: Risk analysis for skilled providers
Output Stage

The 10 Generated Outputs

Every care plan generates 10 tailored documents designed for different audiences and use cases.

1. Caregiver Shift Guide

PRIMARY

The most important document for direct care staff. Contains everything a caregiver needs to know for each shift.

What's Included

  • • Shift-specific task checklists
  • • Timing requirements (meals, meds, etc.)
  • • Communication approach guidance
  • • Safety considerations
  • • What to watch for and when to call

Use Case

Printed and provided to every caregiver before their first shift. Reduces orientation time and ensures consistency across all caregivers.

2. Risk Dashboard

INTERNAL

At-a-glance risk summary with color-coded severity ratings for immediate visibility.

🔴 High Risk

Falls, choking, elopement

🟡 Moderate Risk

UTIs, skin breakdown, isolation

🟢 Monitored

Medication changes, diet

3. Pattern Analysis

INTERNAL

AI-identified behavioral and medical patterns that may not be obvious from individual data points.

Example Patterns Identified

  • • "Multiple recent losses + social isolation + medication non-compliance → likely grief-related"
  • • "Stopped driving + lives alone + decreased church attendance → isolation risk increasing"
  • • "Unsteady + lives alone all day + bathroom far from bedroom → high fall risk during night"

4. Hidden Risks

INTERNAL

Risks not explicitly stated but inferred from assessment data — things humans might miss.

Why This Matters: A client might not mention they skip medications due to cost, or that they avoid showering because they're afraid of falling. AI connects dots across all four forms to surface these unspoken concerns.

5. Strengths & Goals

ALL AUDIENCES

Person-centered focus on what the client CAN do, wants to maintain, and hopes to achieve.

What They Can Still Do

Maintains abilities to preserve

What They Want to Do

Goals that matter to them

6. Escalation Tree

CAREGIVER

Decision flowchart: if you observe X, then do Y. When to document, when to call office, when to call 911.

Example Scenarios

  • Client refuses morning medication → Document time, reason, notify office by end of shift
  • Client appears confused or disoriented → Call office immediately for guidance
  • Client has fallen and cannot get up → Call 911, then notify office

7. Communication Guide

CAREGIVER

How to talk with this specific client: topics they enjoy, things to avoid, preferred communication style.

✅ Conversation Starters

Safe topics that engage them

⚠️ Topics to Avoid

Sensitive subjects or triggers

8. Care Plan Narrative

FORMAL DOCUMENT

Traditional care plan format for chart documentation and external provider communication.

Use Case: Required documentation for agency files. Can be shared with case managers, skilled nursing, or other providers who need formal care plan documentation.

9. Watch List & Medications

CAREGIVER

Quick reference: what to observe, when to be concerned, medication schedule and side effects to monitor.

What Caregivers Watch For

  • • Signs of UTI (confusion, urgency, odor)
  • • Medication side effects (dizziness, drowsiness)
  • • Behavioral changes that signal problems
  • • Skin condition and pressure points

10. Equipment & Environment

INTERNAL

Safety equipment recommendations and home-specific considerations for caregivers.

Equipment Needed

Grab bars, shower bench, walker, raised toilet seat, etc.

Home Safety Notes

Hazards, access instructions, pet considerations

Why 10 Different Documents?

Different audiences need different information. Caregivers need task checklists. Families need reassurance without medical jargon. Administrators need compliance and staffing info. Clinical partners need risk data. One generic care plan can't serve all these needs effectively.

Traditional approach: Write one care plan, spend 2-4 hours, everyone gets the same document.
AI approach: Generate 10 targeted documents in 5 minutes, each optimized for its audience.

Implementation Workflow

Technical documentation of the form chain, code generation, and HIPAA architecture

Form Chain & Code Flow

1

Phone Intake (PHI Form - Step 1)

Staff or client fills out. Contains real name, address, SSN, etc.

Platform: HIPAA-compliant JotForm
Contains PHI: Yes (real names, addresses, contact info)
2

Code Generation (External)

Staff generates unique client code using external tool (e.g., "Gentle Sparrow")

Code stored: Physical paper log in locked cabinet
Digital record: "John Smith = Gentle Sparrow"
Client awareness: Client never knows their code
3

Client Profile (Step 2) - Prefilled Link

Staff sends unique URL: form2.com?code=GentleSparrow

Client clicks link → code auto-populated in hidden field → client fills out form

Platform: Regular JotForm (prefills allowed)
Contains PHI: No (uses code only)
Save & Return: JotForm allows draft saving for long forms
4

Person-Centered Survey (Form 3) - Auto-Redirect

After Step 2 submits → auto-redirect to survey.com?code=GentleSparrow

Code passes through redirect URL → hidden field captures it → client continues

Platform: Regular JotForm
Contains PHI: No (uses code only)
Client experience: Seamless flow, never sees code
5

Functional Assessment (Form 4) - Staff Completes

Staff completes during in-person visit → manually enters client code

Platform: Regular JotForm
Completed by: Staff in-person
Code entry: Manual (staff has code from Step 2)
6

Environmental Assessment (Form 5) - Staff Completes

Staff completes during in-person visit → manually enters client code

Platform: Regular JotForm
Completed by: Staff in-person
Code entry: Manual (staff has code from Step 2)

HIPAA Compliance Architecture

PHI Form (HIPAA Required)

  • Form 1: Phone Intake
  • Platform: HIPAA-compliant JotForm
  • Contains: Real names, addresses, SSN, phone, email
  • Limitation: Cannot use prefills

De-Identified Forms (Regular JotForm)

  • Forms 2-5: All use code only
  • Platform: Regular JotForm (prefills allowed)
  • Contains: Client code, assessment data only
  • Benefit: Can use prefills, redirects, save-and-return

Key Security Feature: The only link between real identity and client code exists on physical paper in a locked cabinet. No digital connection exists between PHI and assessment data.

Technical Implementation Notes

Prefill Chain

Forms 2 & 3 use URL parameter passing: ?code=ClientCode in redirect URLs. JotForm hidden fields capture the parameter automatically.

Save & Return Feature

Since forms are long, enable JotForm's "Save & Continue Later" feature. Clients get a unique link to resume their progress if they take a break.

Code Generation

Use external random name generator or manual assignment. Code must be generated OUTSIDE the PHI form to maintain separation. Staff creates physical log entry before sending Form 2 link.

Small Agency Consideration

The code system adds complexity that may not be necessary for very small agencies. Alternative: use HIPAA JotForm for all forms and skip the code architecture entirely. The prefill approach is more elegant but requires the PHI/de-identification split.