CohortEconomics

  1. Overview
  2. The Rosetta Stone: OMOP to HEOR Mapping
  3. Installation
  4. 6-Stage Pipeline Architecture
  5. Getting Started Example
  6. Core Stages in Detail
    1. Stage 1: Cohort Setup & Initialization (init)
    2. Stage 2: Baseline & HCRU Extraction (summarise_baseline, extract_hcru)
    3. Stage 3: Causal Propensity Score Adjustment (fit_ps, adjust_ps, assess_balance)
    4. Stage 4: Trajectory Compilation & State Costs (compile_trajectories)
    5. Stage 5: Economic Simulation (simulate_economics)
    6. Stage 6: Decision Analysis & CEA (run_cea, plot_*)
  7. Main Functions

The CohortEconomics R package provides an end-to-end analytical framework for Health Economics and Outcomes Research (HEOR), causal inference, Markov state-transition modeling, and Cost-Effectiveness Analysis (CEA) directly on Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) databases.

Overview

Health Technology Assessment (HTA) decision-making requires evaluating whether new clinical interventions provide sufficient health gains relative to their incremental costs. Traditionally, translating observational real-world data (RWD) into decision-analytic models required disconnected, manual data processing steps.

CohortEconomics establishes a standardized, reproducible bridge between OMOP CDM cohorts and decision-analytic modeling engines, enforcing a strict 6-stage pipeline:

  1. Stage 1 (Cohort Initialization): Define Target, Comparator, and Safety/Efficacy Outcome cohorts.
  2. Stage 2 (Baseline & HCRU Characterization): Profile patient demographics and aggregate healthcare resource utilization.
  3. Stage 3 (Causal Propensity Score Adjustment): Control for confounding by indication via regularized logistic regression (Cyclops) and greedy caliper matching.
  4. Stage 4 (Trajectory Compilation & State Costs): Transform longitudinal clinical histories into discrete Markov health-state transition matrices and parametric cost distributions.
  5. Stage 5 (Economic Simulation): Execute Probabilistic Sensitivity Analysis (PSA) Markov microsimulations over defined time horizons with annual discounting.
  6. Stage 6 (Decision Analysis / CEA): Calculate Incremental Cost-Effectiveness Ratios (ICER), Net Monetary Benefit (NMB), Cost-Effectiveness Acceptability Curves (CEAC), and Cost-Effectiveness Planes via BCEA.

The Rosetta Stone: OMOP to HEOR Mapping

CohortEconomics translates OHDSI clinical conventions into health economic evaluation concepts:

OMOP / OHDSI Concept HEOR / CEA Concept Pipeline Stage
Target Cohort (e.g., new users of Drug A) Treatment Arm (The new intervention) Stage 1 (init)
Comparator Cohort (e.g., new users of Drug B) Standard of Care Arm (The baseline comparator) Stage 1 (init)
Outcome Cohorts (e.g., clinical events) Health States / Clinical Endpoints (Markov states) Stage 1 & 4 (init, compile_trajectories)
Baseline Demographics & Covariates Confounders & Patient Profiles (PS Matching) Stage 2 & 3 (summarise_baseline, fit_ps)
COST & VISIT_OCCURRENCE Tables HCRU & Direct Medical Expenditures (ICER numerator) Stage 2 & 4 (extract_hcru, state costs)

Installation

Install CohortEconomics from GitHub:

# install.packages("pak")
pak::pkg_install("iomedhealth/omopHeor/packages/CohortEconomics")

6-Stage Pipeline Architecture 

graph TD
    subgraph "Input Layer"
        A[("OMOP CDM Database")]
    end

    subgraph "Stage 1: Cohort Setup"
        S1["init()<br/>Link Target, Comparator, Outcome Cohorts"]
    end

    subgraph "Stage 2: Baseline & HCRU"
        S2A["summarise_baseline()<br/>Patient Profiles & Demographics"]
        S2B["extract_hcru()<br/>Encounter Volumes & Direct Costs"]
    end

    subgraph "Stage 3: Causal Adjustment"
        S3A["fit_ps()<br/>Regularized Logistic Regression (Cyclops)"]
        S3B["adjust_ps()<br/>Caliper Matching / Weighting"]
        S3C["assess_balance()<br/>SMD & Diagnostic Balance"]
    end

    subgraph "Stage 4: Trajectories & Costs"
        S4A["compile_trajectories()<br/>Discrete Health State Transitions"]
        S4B["Parametric State Costs<br/>Gamma / Log-Normal Distributions"]
    end

    subgraph "Stage 5: Simulation"
        S5["simulate_economics()<br/>Markov PSA State-Transition Model"]
    end

    subgraph "Stage 6: Decision Analysis (CEA)"
        S6["run_cea()<br/>ICER, NMB & BCEA Engine"]
        P1["plot_ceac()<br/>Acceptability Curves"]
        P2["plot_plane()<br/>CE Plane"]
        T1["table_summary()<br/>Executive HTA Summary"]
    end

    A --> S1
    S1 --> S2A --> S2B
    S2B --> S3A --> S3B --> S3C
    S3C --> S4A --> S4B
    S4B --> S5
    S5 --> S6
    S6 --> P1
    S6 --> P2
    S6 --> T1

Getting Started Example

The following end-to-end example demonstrates how to run a complete comparative HEOR analysis using DuckDB and the synthetic GiBleed dataset:

library(CohortEconomics)
library(CDMConnector)
library(CohortConstructor)
library(dplyr)

# 0. Connect to OMOP CDM database
con <- DBI::dbConnect(duckdb::duckdb(), eunomiaDir("GiBleed"))
cdm <- cdmFromCon(con, cdmSchema = "main", writeSchema = "main")

# 1. Instantiate Study Cohorts
cdm$target_cohort <- conceptCohort(cdm, list(celecoxib = 1118084L), "target_cohort") |>
  requireIsFirstEntry()
cdm$comparator_cohort <- conceptCohort(cdm, list(diclofenac = 1124300L), "comparator_cohort") |>
  requireIsFirstEntry()
cdm$outcome_cohort <- conceptCohort(cdm, list(gi_bleed = 192671L), "outcome_cohort")

# 2. Run the 6-Stage Pipeline
study <- init(
  cdm = cdm,
  target_cohort = "target_cohort",
  comparator_cohort = "comparator_cohort",
  outcome_cohort = "outcome_cohort"
) |>
  summarise_baseline() |>
  extract_hcru(
    baseline_window = c(-365, -1),
    followup_window = c(0, 365)
  ) |>
  fit_ps() |>
  adjust_ps(caliper = 0.2) |>
  compile_trajectories()

# 3. Assign State Cost Distributions
study$costs <- data.frame(
  health_state = c("State_Baseline", "State_Outcome"),
  mean_cost = c(450, 4800),
  se_cost = c(40, 350)
)

# 4. Economic Simulation & Decision Analysis
sim <- simulate_economics(
  traj_obj = study,
  time_horizon = 10,
  discount_rate = 0.03,
  n_samples = 250
)

cea <- run_cea(sim)

# 5. Generate HTA Decision Visualizations
plot_ceac(cea)
plot_plane(cea)
table_summary(cea)

Core Stages in Detail

Stage 1: Cohort Setup & Initialization (init)

Validates that the target, comparator, and outcome cohort tables exist in the CDM reference and computes initial cohort record and subject counts via omopgenerics::cohortCount().

Stage 2: Baseline & HCRU Extraction (summarise_baseline, extract_hcru)

  • summarise_baseline(): Computes unadjusted baseline demographic characteristics (age, sex, prior observation).
  • extract_hcru(): Extracts encounter rates and resource volumes across Inpatient, ICU, Emergency, Outpatient, Prescription, and Procedure domains.

Stage 3: Causal Propensity Score Adjustment (fit_ps, adjust_ps, assess_balance)

  • fit_ps(): Fits high-dimensional regularized logistic regression models using Cyclops to estimate individual propensity scores $P(\text{Treatment} \mid X)$.
  • adjust_ps(): Performs greedy nearest-neighbor caliper matching or IPTW weighting to construct balanced comparative cohorts.
  • assess_balance(): Computes pre- and post-adjustment Standardized Mean Differences (SMD) to confirm covariate balance.

Stage 4: Trajectory Compilation & State Costs (compile_trajectories)

  • Slices longitudinal patient histories into discrete Markov health-state transitions (State_Baseline $\rightarrow$ State_Outcome) to derive empirical transition probability matrices.
  • Links state-specific cost distributions (Gamma/Log-Normal) and health state utilities.

Stage 5: Economic Simulation (simulate_economics)

Runs a Markov state-transition Probabilistic Sensitivity Analysis (PSA). At each Monte Carlo iteration:

  • Transition probabilities are drawn from Dirichlet/Gamma distributions.
  • State costs are sampled from Gamma distributions.
  • Health state utilities are sampled from Beta distributions.
  • Future costs and QALYs are accumulated and discounted at a chosen annual rate (e.g., 3%).

Stage 6: Decision Analysis & CEA (run_cea, plot_*)

Wraps BCEA to calculate:

  • Incremental Cost-Effectiveness Ratio (ICER): $\Delta \text{Cost} / \Delta \text{QALY}$.
  • Net Monetary Benefit (NMB): $\text{NMB}(k) = k \cdot \Delta E - \Delta C$ across willingness-to-pay ($k$) thresholds.
  • Visualizations: Cost-Effectiveness Acceptability Curve (plot_ceac) and Cost-Effectiveness Plane (plot_plane).

Main Functions

Stage Function Purpose
Stage 1 init() Initialize a study object linking target, comparator, and outcome cohorts.
Stage 2 summarise_baseline() Generate baseline demographic and comorbidity summary tables.
Stage 2 extract_hcru() / extractHcru() Extract healthcare utilization counts across care domains and link OMOP cost records.
Stage 3 fit_ps() Fit high-dimensional regularized logistic regression for propensity scores using Cyclops.
Stage 3 adjust_ps() Apply greedy caliper matching or weighting to balance comparative cohorts.
Stage 3 assess_balance() Compute Standardized Mean Differences (SMD) and diagnostic balance tables.
Stage 4 compile_trajectories() Build discrete Markov health-state transition matrices from longitudinal patient journeys.
Stage 5 simulate_economics() Run Markov state-transition Probabilistic Sensitivity Analysis (PSA) over a defined time horizon.
Stage 6 run_cea() Execute Bayesian Cost-Effectiveness Analysis (BCEA) and calculate ICER / NMB.
Reporting plot_ceac() Generate the Cost-Effectiveness Acceptability Curve (CEAC) plot.
Reporting plot_plane() Generate the Cost-Effectiveness Plane scatter plot.
Reporting table_summary() Produce an executive summary table of expected costs, QALYs, and ICER.