refactor: fix mart_patient_360 — eliminate cartesian explosion, hash SSN, add financial metrics

models/marts/_marts.yml

@@ -20,3 +20,59 @@ models:
     description: "Provider-level quality metrics"
     config:
       materialized: table
+
+  - name: mart_patient_360
+    description: >
+      Unified patient record aggregating demographics, encounter history,
+      diagnoses, medications, lab results, and financial data into a single
+      patient-grain table. SSN is SHA-256 hashed for HIPAA compliance.
+    config:
+      materialized: table
+      schema: marts_pii
+    meta:
+      contains_pii: true
+      owner: data-governance
+      compliance: HIPAA
+    columns:
+      - name: patient_id
+        description: "Primary key — unique patient medical record number"
+      - name: patient_name
+        description: "Patient full name from registration (PII)"
+      - name: date_of_birth
+        description: "Patient date of birth (PII/PHI)"
+      - name: blood_type
+        description: "Patient blood type"
+      - name: insurance_id
+        description: "Insurance policy identifier"
+      - name: ssn_hash
+        description: "SHA-256 hash of SSN — raw SSN is never exposed downstream"
+      - name: phone
+        description: "Patient phone number (PII)"
+      - name: email
+        description: "Patient email address (PII)"
+      - name: address
+        description: "Patient mailing address (PII)"
+      - name: primary_provider_name
+        description: "Name of the most recent attending provider, derived from stg_providers via encounter history"
+      - name: total_encounters
+        description: "Count of all encounters for this patient"
+      - name: first_encounter_date
+        description: "Date of the patient's earliest encounter"
+      - name: last_encounter_date
+        description: "Date of the patient's most recent encounter"
+      - name: unique_diagnosis_count
+        description: "Count of distinct ICD diagnosis codes across all encounters"
+      - name: primary_diagnosis_codes
+        description: "Comma-separated list of all distinct diagnosis codes"
+      - name: unique_medication_count
+        description: "Count of distinct medications prescribed across all encounters"
+      - name: total_lab_tests
+        description: "Count of all lab tests ordered for this patient"
+      - name: abnormal_lab_count
+        description: "Count of lab results flagged as abnormal"
+      - name: lifetime_charges_dollars
+        description: "Total billed amount across all claims, converted from cents to dollars"
+      - name: lifetime_paid_dollars
+        description: "Total paid amount across all claims, converted from cents to dollars"
+      - name: patient_registered_at
+        description: "Timestamp when the patient record was created"

models/marts/mart_patient_360.sql

@@ -7,69 +7,172 @@
         'owner': 'data-governance',
         'compliance': 'HIPAA'
     }
-) }}with patients as (
+) }}
+
+with patients as (
 
     select * from {{ ref('stg_patients') }}
 
-)
-,
+),
 
 encounters as (
 
     select * from {{ ref('stg_encounters') }}
 
-)
-,
+),
+
+providers as (
+
+    select * from {{ ref('stg_providers') }}
+
+),
 
 diagnoses as (
 
     select * from {{ ref('stg_diagnoses') }}
 
-)
-,
+),
 
 medications as (
 
     select * from {{ ref('stg_medications') }}
 
-)
-,
+),
 
 lab_results as (
 
     select * from {{ ref('stg_lab_results') }}
 
-)
-,
+),
+
+claims as (
+
+    select * from {{ ref('stg_claims') }}
+
+),
+
+-- Pre-aggregate encounters to patient grain to prevent cartesian explosion
+encounter_summary as (
+
+    select
+        enc.patient_id,
+        count(*) as total_encounters,
+        min(enc.admission_date) as first_encounter_date,
+        max(enc.admission_date) as last_encounter_date,
+        -- Get the primary provider from the most recent encounter
+        max(prov.full_name) as primary_provider_name
+    from encounters as enc
+    left join providers as prov
+        on enc.attending_physician_id = prov.provider_id
+    group by enc.patient_id
+
+),
+
+-- Pre-aggregate diagnoses to patient grain
+diagnosis_summary as (
+
+    select
+        enc.patient_id,
+        count(distinct diag.diagnosis_code) as unique_diagnosis_count,
+        listagg(distinct diag.diagnosis_code, ', ') as primary_diagnosis_codes
+    from diagnoses as diag
+    inner join encounters as enc
+        on diag.encounter_id = enc.encounter_id
+    group by enc.patient_id
+
+),
+
+-- Pre-aggregate medications to patient grain
+medication_summary as (
+
+    select
+        enc.patient_id,
+        count(distinct med.medication_name) as unique_medication_count
+    from medications as med
+    inner join encounters as enc
+        on med.encounter_id = enc.encounter_id
+    group by enc.patient_id
+
+),
+
+-- Pre-aggregate labs to patient grain
+lab_summary as (
+
+    select
+        patient_id,
+        count(*) as total_lab_tests,
+        sum(case when abnormal_flag = 'True' then 1 else 0 end) as abnormal_lab_count
+    from lab_results
+    group by patient_id
+
+),
+
+-- Pre-aggregate claims to patient grain using cents_to_dollars macro
+claims_summary as (
+
+    select
+        patient_id,
+        {{ cents_to_dollars('sum(billed_cents)') }} as lifetime_charges_dollars,
+        {{ cents_to_dollars('sum(paid_cents)') }} as lifetime_paid_dollars
+    from claims
+    group by patient_id
+
+),
 
 final as (
 
     select
         -- Primary key
+        patients.patient_id,
+
+        -- Patient demographics
+        patients.full_name as patient_name,
+        patients.date_of_birth,
+        patients.blood_type,
+        patients.insurance_id,
+
+        -- PII columns (HIPAA: hashed for downstream safety)
+        sha2(patients.ssn, 256) as ssn_hash,
+        patients.phone,
+        patients.email,
+        patients.address,
+
+        -- Provider context (derived from encounters + providers)
+        enc_s.primary_provider_name,
+
+        -- Encounter metrics
+        coalesce(enc_s.total_encounters, 0) as total_encounters,
+        enc_s.first_encounter_date,
+        enc_s.last_encounter_date,
+
+        -- Diagnosis metrics
+        coalesce(dx_s.unique_diagnosis_count, 0) as unique_diagnosis_count,
+        dx_s.primary_diagnosis_codes,
 
-        -- Dimensions
+        -- Medication metrics
+        coalesce(med_s.unique_medication_count, 0) as unique_medication_count,
 
+        -- Lab metrics
+        coalesce(lab_s.total_lab_tests, 0) as total_lab_tests,
+        coalesce(lab_s.abnormal_lab_count, 0) as abnormal_lab_count,
 
-        -- PII columns (restricted access)
-        -- HIPAA: These fields require role-based access control
-        full_name,
-        ssn,
-        date_of_birth,
-        phone,
-        email,
-        address,
-        diagnoses.diagnosis_code
+        -- Financial summary
+        coalesce(clm_s.lifetime_charges_dollars, 0) as lifetime_charges_dollars,
+        coalesce(clm_s.lifetime_paid_dollars, 0) as lifetime_paid_dollars,
 
+        patients.created_at as patient_registered_at
 
     from patients
-    left join encounters
-        on patients.patient_id = encounters.patient_id
-    left join diagnoses
-        on encounters.encounter_id = diagnoses.encounter_id
-    left join medications
-        on encounters.encounter_id = medications.encounter_id
-    left join lab_results
-        on patients.patient_id = lab_results.patient_id
+    left join encounter_summary as enc_s
+        on patients.patient_id = enc_s.patient_id
+    left join diagnosis_summary as dx_s
+        on patients.patient_id = dx_s.patient_id
+    left join medication_summary as med_s
+        on patients.patient_id = med_s.patient_id
+    left join lab_summary as lab_s
+        on patients.patient_id = lab_s.patient_id
+    left join claims_summary as clm_s
+        on patients.patient_id = clm_s.patient_id
 
 )