Adaptive Assessment Pattern for GC-SDOH-30

Overview

This document describes a 3-tiered adaptive assessment pattern that reduces assessment time by 60%+ for users with low-risk profiles while maintaining data quality for those who need more detailed evaluation.

Problem Statement

While comprehensive assessments like the GC-SDOH-30 provide valuable insights, they can be time-consuming for users who have stable conditions across most domains. The adaptive pattern addresses this by:

Starting with a quick screening for returning users
Targeting detailed questions only to areas that show elevated need
Maintaining comprehensive baseline assessments for initial evaluations

Architecture: 3-Tiered Progressive Assessment

Tier 1: SDOH-Quick-6 (2 minutes)

Purpose: Fast screening to identify problem zones

Structure:

6 questions total (1 per priority zone)
Uses highest item-total correlation questions from each zone
Default starting point for returning users

Question Selection Criteria: Select the question from each zone that has the highest correlation with the total zone score from validation data:

P1 (Social Support): Choose question with strongest correlation to social support needs
P2 (Physical Health): Choose question with strongest correlation to physical wellbeing
P3 (Housing): Choose question with strongest correlation to housing stability
P4 (Financial): Choose question with strongest correlation to financial resources
P5 (Legal/Navigation): Choose question with strongest correlation to navigation ability
P6 (Emotional): Choose question with strongest correlation to emotional wellbeing

Scoring:

Calculate preliminary zone scores (1 data point per zone, confidence = 1/N where N is total questions in zone)
Flag zones with scores >50 (on normalized 0-100 scale) for deep dive

Tier 2: SDOH-Deep-Dive (3-4 minutes)

Purpose: Targeted detail gathering for flagged zones

Structure:

Dynamic question set based on Quick-6 results
3-4 questions per flagged zone only
Skips zones that scored <50 in Quick-6

Question Selection: For each flagged zone, select 3-4 additional questions that:

Weren't asked in Quick-6
Provide maximum additional information (highest item-remainder correlation)
Cover different aspects of the zone

Example Deep-Dive Sets:

P1 Deep Dive (4 questions):

Questions 2, 4, 6, 8 from original P1 set (if Q1 was used in Quick-6)

P2 Deep Dive (2 questions):

Both questions (Q29, Q30) if not already in Quick-6

P3 Deep Dive (3 questions):

Questions covering stability, safety, and adequacy

P4 Deep Dive (4 questions):

Questions covering healthcare costs, basic needs, utilities, transportation

P5 Deep Dive (3 questions):

Questions covering communication, coordination, legal preparedness

P6 Deep Dive (2 questions):

Both questions if not already in Quick-6

Scoring:

Combine Quick-6 and Deep-Dive responses
Calculate updated zone scores with higher confidence
Generate zone-specific recommendations

Tier 3: SDOH-30 Comprehensive (5-6 minutes)

Purpose: Complete baseline assessment and periodic comprehensive evaluation

Structure:

All 30 questions from GC-SDOH-30
Used for initial assessments (need baseline)
Recommended monthly for ongoing monitoring

When to Use:

First-time assessment (always)
Monthly comprehensive check-in
After major life changes
When user requests full assessment

Decision Logic

START
│
├─ First assessment?
│  ├─ YES → SDOH-30 Comprehensive
│  └─ NO → Continue
│
├─ Monthly check-in due?
│  ├─ YES → Offer SDOH-30 Comprehensive
│  └─ NO → Continue
│
├─ Major life change reported?
│  ├─ YES → Offer SDOH-30 Comprehensive
│  └─ NO → Continue
│
└─ Regular check-in:
   │
   ├─ Step 1: Administer SDOH-Quick-6
   │
   ├─ Step 2: Calculate zone scores
   │
   ├─ Step 3: Any zone score >50?
   │  │
   │  ├─ YES → Offer Deep Dive
   │  │   │
   │  │   ├─ "I see [zone names] could use more attention."
   │  │   ├─ "May I ask 3-4 more questions about these areas
   │  │   │   for better recommendations?"
   │  │   │
   │  │   ├─ User accepts?
   │  │   │  ├─ YES → Administer Deep Dive for flagged zones
   │  │   │  └─ NO → Complete with Quick-6 data only
   │  │   │
   │  │   └─ Calculate final scores
   │  │
   │  └─ NO → Complete (all zones <50)
   │      │
   │      └─ "Great news! Everything looks stable today."

Implementation Considerations

Data Quality

Confidence Scoring:

Track how many questions were answered per zone
Store confidence level with each zone score
Example: P1 with 1/8 questions = 12.5% confidence

Score Adjustment:

Quick-6 only: Lower confidence, wider margin of error
Deep-Dive: Higher confidence for specific zones
Comprehensive: Highest confidence across all zones

Validation: Recommend parallel testing period:

Administer both Quick-6 and full SDOH-30 to sample population
Compare zone scores from Quick-6 vs. full assessment
Validate that Quick-6 correctly identifies high-risk zones (sensitivity >85%)
Ensure low false-negative rate (<10%)

User Experience

Messaging:

Quick-6 Introduction: "Quick check-in time! Just 6 questions today."
Deep-Dive Offer: "I notice [zones] might need support. May I ask a few more questions for better recommendations?"
Optional: Always make Deep-Dive optional, frame as benefit
Completion: "All done! Based on your responses, here are some resources..."

Timing:

Quick-6: ~30 seconds per question = 3 minutes total (with reading time)
Deep-Dive: ~30-45 seconds per question = 2-3 minutes for 4 questions
Total time saved: 60-70% for low-risk users

Privacy & Consent

Informed Consent:

Explain adaptive assessment approach
Clarify that fewer questions may mean less detailed recommendations
Allow user to request full assessment at any time

Data Storage:

Link Deep-Dive responses to originating Quick-6 assessment
Store assessment chain for audit trail
Track which tier was used for each assessment

Success Metrics

Completion Rates

Target: >85% (vs. ~70% for full SDOH-30)
Track by tier: Quick-6, Deep-Dive acceptance, Full completion

Time Efficiency

Average time saved per assessment
Target: 60%+ reduction for users scoring <50 on all zones

Data Quality

Zone score correlation: Quick-6 vs. Full assessment
Target: r >0.85 for zone identification
False negative rate: <10%

User Satisfaction

Survey: "The assessment length felt appropriate" (agree/strongly agree)
NPS score for assessment experience
Voluntary full assessment requests (should be <15%)

Risk Mitigation

Risk 1: Quick-6 Misses High-Need Zone

Mitigation:

Validate question selection using item-total correlation from pilot data
Choose questions with highest predictive power
Monitor false-negative rate in production
Adjust threshold if needed (lower from 50 to 45)

Risk 2: Users Skip Deep-Dive

Mitigation:

Frame as "better recommendations" not "more burden"
Keep truly optional
Track Deep-Dive acceptance rate
If <50%, revise messaging or auto-trigger for very high scores (>70)

Risk 3: Score Inconsistency

Mitigation:

Run parallel testing for 2-4 weeks
Document expected confidence intervals by tier
Provide score range (±5 points) for Quick-6 results
Recommend full assessment for borderline cases

Risk 4: Loss of Longitudinal Data

Mitigation:

Ensure monthly comprehensive assessments
Store all historical assessment data
Calculate trend scores using time-weighted averaging
Alert if user hasn't completed comprehensive in >60 days

Implementation Roadmap

Phase 1: Assessment Definition (Week 1)

Define SDOH-Quick-6 question set
Create Deep-Dive question pools per zone
Update assessment validators

Phase 2: Decision Logic (Week 1-2)

Implement zone risk detection
Create Deep-Dive triggering logic
Build question selection algorithms

Phase 3: Scoring System (Week 2)

Implement confidence scoring
Handle partial zone data
Merge Quick-6 and Deep-Dive results

Phase 4: Testing & Validation (Week 2-3)

Unit tests for all scoring logic
Integration tests for assessment flows
Parallel testing with sample users

Phase 5: Rollout (Week 3-4)

Deploy with feature flag
Beta test with 5-10 users
Monitor metrics
Gradual rollout to 25%, then 100%

Technical Specifications

Assessment Type Identifiers

sdoh_quick: Quick-6 screening
sdoh_deep: Deep-Dive targeted assessment
sdoh: Full SDOH-30 comprehensive

Data Schema Requirements

Assessments Table:

type: Assessment type identifier
parentAssessmentId: Links Deep-Dive to Quick-6
zonesFlagged: Array of zone IDs that triggered Deep-Dive

Scores Table:

instrument: Assessment type used
confidence: Confidence level per zone (0-1)
answeredRatio: Questions answered / total questions per zone

Assessment Sessions Table:

assessmentType: Current tier being administered
deepDiveZones: Zones requiring Deep-Dive
deepDiveOffered: Boolean
deepDiveAccepted: Boolean

Scoring Formulas

Zone Score (Quick-6):

score = (answer - 1) / 4 * 100  // Convert 1-5 to 0-100
confidence = 1 / totalQuestionsInZone

Zone Score (Deep-Dive):

quickAnswer = quickScreenResponses[zone]
deepAnswers = deepDiveResponses[zone]
allAnswers = [quickAnswer, ...deepAnswers]
score = mean(allAnswers.map(a => (a - 1) / 4 * 100))
confidence = allAnswers.length / totalQuestionsInZone

Risk Threshold:

if (zoneScore > 50 && confidence < 0.5) {
  flagForDeepDive = true
}

Example User Flows

Flow 1: Low-Risk User

User starts check-in
System: "Quick 6-question check-in today!"
User completes Quick-6 (2 minutes)
All zones score <50
System: "Great! Everything looks stable. Here are your resources."
Time saved: 3-4 minutes

Flow 2: Mixed-Risk User

User starts check-in
User completes Quick-6 (2 minutes)
P4 (Financial) scores 65, P5 (Legal) scores 55
System: "I notice financial resources and healthcare navigation might need support. May I ask 4 more questions for better recommendations?"
User accepts
User answers 4 Deep-Dive questions for P4 and P5 (2 minutes)
System provides targeted financial and navigation resources
Time saved: 1-2 minutes vs. full assessment

Flow 3: High-Risk User

User starts monthly check-in
System: "It's time for your monthly comprehensive assessment"
User completes all 30 questions (5-6 minutes)
Multiple zones flagged
System connects user to case management
Full data quality maintained for high-need user

References

GC-SDOH-30 Assessment Framework
Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences (PRAPARE)
Accountable Health Communities (AHC) screening tool
Computer Adaptive Testing (CAT) methodologies

Version History

v1.0 (2025-11-23): Initial adaptive assessment pattern documentation

For implementation guidance specific to your tech stack, see platform-specific documentation.

FilesExpand file tree

ADAPTIVE-ASSESSMENT-PATTERN.md

Latest commit

History

ADAPTIVE-ASSESSMENT-PATTERN.md

File metadata and controls

Adaptive Assessment Pattern for GC-SDOH-30

Overview

Problem Statement

Architecture: 3-Tiered Progressive Assessment

Tier 1: SDOH-Quick-6 (2 minutes)

Tier 2: SDOH-Deep-Dive (3-4 minutes)

Tier 3: SDOH-30 Comprehensive (5-6 minutes)

Decision Logic

Implementation Considerations

Data Quality

User Experience

Privacy & Consent

Success Metrics

Completion Rates

Time Efficiency

Data Quality

User Satisfaction

Risk Mitigation

Risk 1: Quick-6 Misses High-Need Zone

Risk 2: Users Skip Deep-Dive

Risk 3: Score Inconsistency

Risk 4: Loss of Longitudinal Data

Implementation Roadmap

Phase 1: Assessment Definition (Week 1)

Phase 2: Decision Logic (Week 1-2)

Phase 3: Scoring System (Week 2)

Phase 4: Testing & Validation (Week 2-3)

Phase 5: Rollout (Week 3-4)

Technical Specifications

Assessment Type Identifiers

Data Schema Requirements

Scoring Formulas

Example User Flows

Flow 1: Low-Risk User

Flow 2: Mixed-Risk User

Flow 3: High-Risk User

References

Version History