Skip to contents

HCRU Extraction Logic

Source: vignettes/hcru_logic.Rmd
hcru_logic.Rmd

HCRU & Direct Cost Extraction Logic

This document outlines the extraction and linkage logic for Healthcare Resource Utilization (HCRU) and direct medical expenditures from the OMOP CDM COST table across the omopHeor ecosystem.

1. Dual Cost Extraction Modalities

omopHeor provides two complementary mechanisms for extracting cost data:

  1. In-Database Cohort Enrichment (CohortCosts::addCosts()):
    • Appends windowed domain cost columns (cost_inpatient_*, cost_outpatient_*, cost_drug_*, cost_procedure_*, cost_total_*) directly to any OMOP cohort table in the database write schema.
    • Suited for epidemiological studies, patient profiling, and multivariable regression.
  2. Pipeline State-Cost Extraction (CohortEconomics::extract_hcru()):
    • Links clinical event expenditures to study subjects and tags each cost with the patient’s health state (e.g. State_Baseline vs State_Outcome).
    • Feeds state-specific cost distributions directly into Stage 4 (compile_trajectories()) and Stage 5 Markov simulations (simulate_economics()).

2. Extraction Decision Flow 

                            +-------------------------------------+
                            | addCosts() or extract_hcru() called |
                            +------------------+------------------+
                                               |
                                               v
                                   +-----------------------+
                                   |   Does 'cost' table   |
                                   |   exist in the CDM?   |
                                   +-------+-------+-------+
                                           |       |
                                      [No] |       | [Yes]
                                           v       v
                        +--------------------+   +---------------------------+
                        | Warn user: No cost |   |   Is 'cost' table empty?  |
                        | table found. Fill  |   |   (record count == 0)     |
                        | columns with 0.0.  |   +-------+---------+---------+
                        +--------------------+           |         |
                                                    [Yes]|         | [No]
                                                         v         |
                                         +--------------------+    |
                                         | Warn user: Table   |    |
                                         | exists but empty.  |    |
                                         | Return 0.0 values. |    |
                                         +--------------------+    v
                                                     +--------------------------------+
                                                     | Are financial metrics          |
                                                     | (total_paid, total_charge)     |
                                                     | available and populated?       |
                                                     +---------+-------------+--------+
                                                               |             |
                                                          [No] |             | [Yes]
                                                               v             |
                                         +---------------------------+       |
                                         | Do drg_concept_id or      |       |
                                         | drg_source_value exist    |       |
                                         | in the table?             |       |
                                         +-------+---------+---------+       |
                                                 |         |                 |
                                            [No] |         | [Yes]           |
                                                 v         v                 |
                          +------------------------+  +-------------------+  |
                          | Zero-fill cost columns |  | Prompt user for   |  |
                          | or warn missing metric |  | DRG-to-Cost       |  |
                          |                        |  | lookup dictionary |  |
                          +------------------------+  +--------+----------+  |
                                                               |             |
                                                               |             |
                                                               v             v
                                                    +--------------------------------+
                                                    | Link by cost_event_id across   |
                                                    | Condition, Visit, Drug, and    |
                                                    | Procedure occurrence tables    |
                                                    +--------------------------------+
                                                               |
                                                               v
                                                    +--------------------------------+
                                                    | Aggregate across temporal      |
                                                    | windows & zero-fill subjects   |
                                                    +--------------------------------+

3. Polymorphic Event Linkage Strategy

OMOP CDM v5.3+ stores costs in a polymorphic COST table where cost_event_id references the primary key of the domain table indicated by cost_domain_id:

SELECT 
  c.person_id,
  c.cost_domain_id,
  c.cost_type_concept_id,
  c.total_paid,
  c.total_charge,
  c.paid_by_payer,
  c.paid_by_patient,
  v.visit_concept_id,
  v.visit_start_date,
  v.visit_end_date
FROM cdm.cost c
JOIN cdm.visit_occurrence v 
  ON c.cost_event_id = v.visit_occurrence_id 
 AND c.cost_domain_id = 'Visit'
WHERE c.person_id IN (SELECT subject_id FROM study_cohort);

Supported Cost Domains

  • Inpatient Visits (cost_domain_id = 'Visit' and visit_concept_id in Inpatient/ICU concepts).
  • Outpatient Visits (cost_domain_id = 'Visit' and visit_concept_id in Outpatient concepts).
  • Pharmacy (cost_domain_id = 'Drug' linked via drug_exposure_id).
  • Procedures & Diagnostics (cost_domain_id = 'Procedure' or 'Measurement').
  • Conditions (cost_domain_id = 'Condition' linked via condition_occurrence_id).

4. Resilience & Fallback Rules

  1. Missing Cost Table: If the database schema lacks a COST table, a clear warning is emitted (Missing 'cost' table in CDM), and all cost columns are safely populated with 0.0 to prevent pipeline crashes.
  2. Empty Cost Table: If the table exists with 0 rows, subjects are retained with 0.0 expenditures.
  3. Missing Subject Records: Zero-utilization subjects in the cohort are preserved via left joins with coalesce to 0.0, ensuring no patient attrition during cost enrichment.