Fall Risk

============================================================ PHASE 2: SENSOR DATA INTEGRATION ANALYSIS

DATA COLLECTION:

• Examine each sensor integration for data completeness and reliability.
• Check for missing data handling: sensor offline, low battery, out of range.
• Verify sensor calibration and validation routines.
• Look for data deduplication and timestamp synchronization across sensors.

SIGNAL PROCESSING:

• Examine noise filtering algorithms: low-pass, Kalman, median filters.
• Check for activity classification from raw accelerometer data: walking, standing, sitting, lying, falling.
• Verify handling of sensor drift and calibration degradation.
• Look for edge computing vs. cloud processing architecture decisions.

FEATURE EXTRACTION:

• Document all derived features used in risk scoring:
  • Gait metrics: stride length, cadence, symmetry, variability, speed.
  • Balance metrics: sway amplitude, center of pressure movement.
  • Activity patterns: daily step count, sedentary time, sleep quality, nocturia frequency.
  • Vital signs: resting heart rate trends, blood pressure patterns, orthostatic changes.
• Verify feature extraction handles missing sensor inputs gracefully.
• Check for temporal windowing: rolling averages, time-of-day normalization.

DATA QUALITY:

• Check for anomaly detection on incoming sensor data: stuck values, impossible readings.
• Verify data validation rules before features enter the risk model.
• Look for data completeness thresholds: minimum sensor uptime required for valid scoring.
• Examine historical data backfill and correction capabilities.

============================================================ PHASE 3: RISK SCORING ALGORITHM ANALYSIS

RISK MODEL ARCHITECTURE:

• Read the risk scoring algorithm in full.
• Identify model type: rule-based scoring, logistic regression, random forest, neural network, ensemble, or hybrid.
• Document all input features and their weights or importance rankings.
• Check for model versioning and A/B testing capabilities.

SCORING FACTORS:

• Intrinsic factors: age, fall history, diagnosis codes, cognitive status, mobility aids used, vision impairment, continence status.
• Behavioral factors: gait changes, activity level changes, sleep disruption, medication changes, missed meals, social isolation.
• Environmental factors: lighting conditions, floor surfaces, clutter detection, bathroom accessibility, stair usage frequency.
• Medication factors: polypharmacy count, high-risk medications (sedatives, antihypertensives, diuretics, psychotropics), recent medication changes.

RISK STRATIFICATION:

• Document risk tiers (e.g., low, moderate, high, critical) and their thresholds.
• Check for dynamic thresholds that adapt to individual baseline patterns.
• Verify that risk scores include confidence intervals or uncertainty measures.
• Look for trend-based scoring: risk increasing over time vs. point-in-time score.

MODEL VALIDATION:

• Check for sensitivity and specificity metrics on historical fall data.
• Look for false positive and false negative rate tracking.
• Examine model retraining pipelines and data drift detection.
• Verify the model handles new residents with limited baseline data.

============================================================ PHASE 4: ENVIRONMENTAL HAZARD IDENTIFICATION

HAZARD DETECTION:

• Examine environmental assessment data models and capture workflows.
• Check for automated hazard detection via sensors: wet floor alerts, poor lighting, obstacle detection from motion patterns.
• Look for structured environmental assessment checklists in the system.
• Verify hazard data feeds into the risk scoring model.

HAZARD CATEGORIES:

• Evaluate coverage of standard fall hazard categories:
  • Lighting: inadequate hallway, bathroom, stairwell lighting.
  • Flooring: loose rugs, wet surfaces, uneven transitions, high-gloss finishes.
  • Furniture: unstable chairs, missing grab bars, bed height.
  • Obstacles: clutter, cords, thresholds, steps without contrast markings.
  • Bathroom: lack of grab bars, slippery tubs, toilet height.
• Check for room-by-room hazard mapping.

REMEDIATION TRACKING:

• Look for hazard remediation workflows: identified, assigned, in progress, resolved.
• Check for re-assessment scheduling after remediation.
• Verify unresolved hazards increase the resident's risk score.
• Examine hazard trend reporting across the facility or home.

============================================================ PHASE 5: MEDICATION INTERACTION ANALYSIS

MEDICATION DATA:

• Examine how medication lists are captured and maintained.
• Check for integration with pharmacy systems or medication dispensing devices.
• Verify medication changes trigger risk score recalculation.
• Look for PRN (as-needed) medication tracking and its impact on scoring.

FALL-RISK MEDICATION FLAGGING:

• Check for a medication risk database or classification system.
• Verify high-risk drug classes are flagged: benzodiazepines, opioids, antihypertensives, diuretics, antipsychotics, antidepressants, anticonvulsants, muscle relaxants, alpha-blockers.
• Look for polypharmacy scoring (risk increases with 4+ medications).
• Check for recent medication change detection: new starts, dose changes, discontinuations.

INTERACTION DETECTION:

• Examine drug-drug interaction checking capabilities.
• Check for additive sedation or hypotension risk detection across multiple medications.
• Verify interaction alerts include fall-specific risk context.
• Look for time-based interaction analysis: multiple sedating medications at bedtime.

============================================================ PHASE 6: MOBILITY TREND TRACKING

LONGITUDINAL ANALYSIS:

• Examine how mobility metrics are stored and trended over time.
• Check for baseline establishment per individual.
• Verify decline detection compares current metrics to personal baseline, not just population averages.
• Look for rate-of-change analysis: gradual decline vs. sudden change.

TREND INDICATORS:

• Check for tracking of: daily step count trends, gait speed changes over weeks, balance test score progression, time-to-stand from seated changes, grip strength measurements, functional reach changes.
• Verify trends account for normal daily variation vs. meaningful decline.
• Look for seasonality adjustments (less activity in winter may be normal).

DECLINE DETECTION:

• Examine threshold logic for flagging meaningful mobility decline.
• Check for statistical significance testing on trend changes.
• Verify decline alerts differentiate acute events (illness, injury) from chronic progression.
• Look for automated care plan recommendations based on mobility trends.

============================================================ PHASE 7: ALERT AND RESPONSE WORKFLOW

ALERT GENERATION:

• Map all alert types: fall detected, high risk score, risk score increase, mobility decline, hazard detected, sensor offline, medication risk change.
• Document severity levels and escalation rules for each alert type.
• Check for alert fatigue mitigation: suppression of low-value repeat alerts, bundling related alerts, quiet hours.
• Verify alert thresholds are configurable per resident or facility.

ALERT ROUTING:

• Examine routing rules: who receives which alerts (nurse, aide, family, physician, emergency services).
• Check for acknowledgment tracking and escalation on non-response.
• Verify alert delivery uses appropriate urgency channels: push notification, SMS, phone call, in-app, pager.
• Look for on-call schedule integration for after-hours alerts.

RESPONSE WORKFLOW:

• Check for guided response protocols attached to alert types.
• Examine post-fall assessment workflows: injury check, vitals, incident report.
• Verify response times are tracked: alert sent to acknowledged to resolved.
• Look for root cause documentation when falls occur.

POST-FALL ANALYSIS:

• Check for post-fall review workflows that update the risk model.
• Verify fall events are correlated with recent risk scores and alerts.
• Look for pattern analysis across falls: time of day, location, activity, medication.
• Examine whether post-fall findings feed back into prevention strategies.

============================================================ SELF-HEALING VALIDATION (max 2 iterations)

After producing output, validate data quality and completeness:

1. Verify all output sections have substantive content (not just headers).
2. Verify every finding references a specific file, code location, or data point.
3. Verify recommendations are actionable and evidence-based.
4. If the analysis consumed insufficient data (empty directories, missing configs), note data gaps and attempt alternative discovery methods.

IF VALIDATION FAILS:

• Identify which sections are incomplete or lack evidence
• Re-analyze the deficient areas with expanded search patterns
• Repeat up to 2 iterations

IF STILL INCOMPLETE after 2 iterations:

• Flag specific gaps in the output
• Note what data would be needed to complete the analysis

============================================================ OUTPUT

Fall Risk Prediction System Analysis

Platform: {detected stack and integrations}

Scope: {subsystems analyzed}

Sensor Types: {N} integrated

Risk Model: {type and version}

Alert Types: {N} configured

System Health Summary

Critical Findings

1. {FALL-001}: {title}
   • Domain: {Sensor/RiskModel/Environment/Medication/Mobility/Alert}
   • Location: {file:line}
   • Impact: {what could go wrong for resident safety}
   • Recommendation: {specific improvement}

Sensor Architecture

• Sensor types: {list}
• Data pipeline: {streaming/batch/hybrid}
• Missing data handling: {present/absent}
• Edge processing: {present/absent}

Risk Model Profile

• Model type: {rule-based/ML/ensemble/hybrid}
• Input features: {N} total
• Risk tiers: {list}
• Validation metrics: {sensitivity/specificity if available}
• Individual baseline: {present/absent}

Environmental Assessment

• Hazard categories covered: {N} of standard set
• Automated detection: {present/absent}
• Remediation tracking: {present/absent}

Medication Integration

• High-risk drug flagging: {present/absent}
• Polypharmacy scoring: {present/absent}
• Interaction checking: {present/absent}
• Change detection: {present/absent}

Alert Effectiveness

• Alert types: {N}
• Fatigue mitigation: {present/absent}
• Escalation on non-response: {present/absent}
• Response time tracking: {present/absent}

DO NOT:

• Do NOT recommend specific medical devices or sensor hardware brands.
• Do NOT make clinical recommendations about medication changes or care interventions.
• Do NOT evaluate the clinical validity of fall risk assessment scales (focus on system implementation).
• Do NOT ignore privacy implications of continuous sensor monitoring.
• Do NOT skip medication interaction analysis even if the system does not currently integrate pharmacy data.
• Do NOT assess staffing adequacy or care quality beyond what the system measures.

NEXT STEPS:

• "Run /medication-adherence to analyze the medication management system in depth."
• "Run /caregiver-coordination to evaluate how fall alerts integrate into care workflows."
• "Run /security-review to audit access controls on sensitive health monitoring data."
• "Run /load-test to validate alert pipeline throughput under high-sensor-count conditions."

============================================================ SELF-EVOLUTION TELEMETRY

After producing output, record execution metadata for the /evolve pipeline.

Check if a project memory directory exists:

• Look for the project path in ~/.claude/projects/
• If found, append to skill-telemetry.md in that memory directory

Entry format:

### /fall-risk — {{YYYY-MM-DD}}
- Outcome: {{SUCCESS | PARTIAL | FAILED}}
- Self-healed: {{yes — what was healed | no}}
- Iterations used: {{N}} / {{N max}}
- Bottleneck: {{phase that struggled or "none"}}
- Suggestion: {{one-line improvement idea for /evolve, or "none"}}

Only log if the memory directory exists. Skip silently if not found. Keep entries concise — /evolve will parse these for skill improvement signals.