UDM Architecture Flow

From Incompatible Sources to Actionable Insights

1
The Problem: Incompatible Source Systems
Column-based structures with different identifiers, scales, and care models
RESILIENCE LAB RL_EPISODE_OUTCOMES
Field Sample Value
patient_id RL-10234
episode_id E-2024-001
care_team_id TEAM-A
phq9_baseline 18.0
phq9_final 6.0
outcome_status Improved
✗ Uses care team ORGANIZATION
✗ PHQ-9 scale (0-27)
✗ TEXT status values
CEREBRAL CEREBRAL_PATIENT_EPISODES
Field Sample Value
member_id C-50012
episode_number 1
practitioner_id PRAC-301
depression_score_initial 72.0
depression_score_final 45.0
treatment_result IMP
✗ Uses individual PERSON
✗ Proprietary scale (0-100)
✗ CODE status values
↓
2
The Solution: Universal Data Model Integration
Combined ERD + Instance Model showing schema relationships with actual data
PARTY
party_idtype
P-RL-10234PERSON
P-C-50012PERSON
TEAM-AORG
PRAC-301PERSON
EPISODE_OUTCOME
outcome_idvalue
OUT-001-B18.0
OUT-001-F6.0
OUT-006-B19.4
OUT-006-F12.2
OUTCOME_TYPE
typeminmax
PHQ9027
EPISODE
episode_idstartend
EP-0012024-01-152024-07-15
EP-0062024-05-012024-11-01
PERSON
party_idname
P-RL-10234J. Smith
P-C-50012L. Garcia
PRAC-301Dr. Anderson
ORGANIZATION
party_idname
TEAM-ARL Team A
TEAM-CEREBRAL-1Cerebral T1
PARTY_ROLE
party_role_idparty_idrole
PR-001P-RL-10234PATIENT
PR-002P-C-50012PATIENT
PR-003TEAM-ACARE_TEAM
PR-004PRAC-301PRACTITIONER
EPISODE_ROLE
episode_role_idepisode_idparty_role_id
ER-001EP-001PR-001
ER-002EP-006PR-002
ER-003EP-001PR-003
ER-004EP-006PR-004
PARTY_RELATIONSHIP
fromtotype
ORG-RL-INCTEAM-AROLLUP
PRAC-CEREBRALPRAC-301AFFIL
VISIT
visit_iddate
VISIT-0012024-04-15
VISIT-006A2024-06-15
VISIT-006B2024-09-15
OFFERING
offering_idname
OFFER-RL-ICRL Integrated Care
OFFER-STDStandard Therapy
DELIVERY
delivery_idvisit_idoffering_id
DEL-001VISIT-001OFFER-RL-IC
DEL-006AVISIT-006AOFFER-STD
DEL-006BVISIT-006BOFFER-RL-IC
DELIVERY_ROLE
delivery_idparty_idrole
DEL-001TEAM-ACARE_TEAM
DEL-006APRAC-301PRACT
DEL-006BTEAM-CEREBRAL-1CARE_TEAM
✓ Patient identifiers unified through PARTY
✓ Scores normalized to PHQ-9 standard
✓ Care teams & practitioners both as PARTY
✓ Methodology tracked via OFFERING
↓
3
The Simplification: Data Mart Aggregation
UDM complexity hidden behind clean dimensional model for business consumption
MONTHLY_TIME_SERIES_FACT
month_key year_month avg_phq9_score episode_count adoption_pct
202412 2024-12 12.30 53 0
202501 2025-01 11.80 54 5
202502 2025-02 11.20 56 15
... ... ... ... ...
202512 2025-12 6.80 78 98
SELECT
  TO_CHAR(episode_start_month, 'YYYYMM')::INT AS month_key,
  episode_start_month AS year_month,
  AVG(final_score) AS avg_phq9_score,
  COUNT(*) AS episode_count,
  MAX(adoption_pct_at_start) AS adoption_pct
FROM episode_fact
GROUP BY episode_start_month
ORDER BY month_key
↓
4
The Answer: Board-Ready Insights
Simple query on data mart answers complex business question
Board Question: Are Cerebral outcomes converging toward Resilience Lab baseline after methodology training?
r = -0.92
Strong correlation between adoption and outcome improvement
Strategic Value Demonstrated:
• Incompatible sources integrated through UDM (not custom ETL)
• MDM as architectural consequence (not separate $20M project)
• Board question answered with simple SQL (not analyst bottleneck)
• Dose-response proves causality (not just correlation)
• Scales to any future integration (acquisitions, partnerships, wearables)