CohortMethod R package
CohortMethod is a software tool that helps researchers compare the effects of different treatments using real-world patient data stored in a standard format. It focuses on studies that look at people who are starting a treatment for the first time, and it uses advanced statistics to reduce bias when analyzing outcomes like disease progression or side effects.
At a glance
Use when
Comparing clinical outcomes between treatments using observational databases like EHRs or claims data, especially when studying new users and needing to adjust for many confounders in a standardized way.
Avoid when
When data is not available in or cannot be mapped to the OMOP CDM, when studying prevalent users of treatments, or when unmeasured confounding is likely to dominate results.
Inputs
Observational health data mapped to the OMOP Common Data Model, treatment and outcome definitions, study design parameters (e.g., time-at-risk, covariates), and database connection details.
Outputs
Propensity scores, adjusted cohort sets (via matching/weighting), covariate balance diagnostics, and effect estimates (e.g., hazard ratios, incidence rate ratios) with confidence intervals and p-values.
How it works
CohortMethod is an R package designed to conduct new-user cohort studies in observational databases structured under the OMOP Common Data Model. It extracts patient-level data, constructs propensity scores using large-scale regularized regression with thousands of covariates (e.g., drugs, diagnoses, procedures, demographics), and supports adjustment via matching, stratification, trimming, or weighting. Outcome models include conditional logistic, Poisson, and Cox regression. The package includes diagnostic tools for assessing covariate balance and propensity score distributions, with core functions implemented in C++ for performance. It requires R ≥ 4.0.0 and Java.
- Project
- EHDEN
- Funding
- IMI
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Categories
- RWEPredictive Modelling
- Technology
- Non-specific
- Assumptions
- Patients are new users of the treatment, exposure periods are well-defined, outcomes are accurately recorded, and sufficient data exists to model confounders; assumes no unmeasured confounding after adjustment.
- Strengths
- Uses comprehensive covariate sets from standardized data models, supports robust confounding adjustment, includes diagnostic tools for balance assessment, scalable to large datasets, and interoperable with the OHDSI analytics ecosystem.
- Limitations
- Requires data to be transformed into the OMOP CDM format, limited to new-user designs, assumes correct specification of propensity and outcome models, and cannot account for unmeasured confounders.
- Also known as
- CohortMethod, OHDSI CohortMethod
Questions this answers
- › How can I compare the effectiveness of two treatments using real-world data from electronic health records?
- › How do I reduce selection bias when studying treatment outcomes in observational databases?
- › Can I analyze time-to-event outcomes like hospitalization or death using this tool?
- › How do I check if my patient groups are balanced after adjusting for confounders?
- › Does this tool support large-scale data with many patient characteristics?
- › Is this method suitable for studying new users of a drug?
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.