Characterization

Nathan Hall and Jenna Reps

2024-10-22

Introduction

The OHDSI Characterization package lets users extract descriptive analyses from observational healthcare data sets mapped to the OMOP CDM. There are currently four different types of characterizations analyses (incidence rates, time-to-event, dechallenge-rechallenge and various aggregate covariate cohort comparisons).

The Characterization package currently lets users answer the following questions:

• Incidence Rate: How often does <add outcome> occur within <add time-at-risk> after first record of <add exposure/indication>?
• Time-to-event: When does <add outcome> occur relative to the first recorded of <add exposure/indication>? Is it more common before or after <add exposure>?
• Dechallenge-rechallenge: Is there any evidence of <add outcome> causing <add exposure> to be discontinued and then <add outcome> re-occurring once <add exposure> restarts?
• Cohort Comparison: What is different at index between patients in <add exposure/outcome/indication> and patients in <add different exposure/outcome/indication>?
• Database Comparison: What is different at index between patients in <add exposure/outcome/indication> across two or more OMOP CDM databases?
• Risk factors: What are the risk factors of <add outcome> occurring within <add time-at-risk> for those exposed to <add exposure>?
• Case-series: What happens to cases (those exposure to <add exposure> who have <add outcome> during <add time-at-risk>) before exposure, between exposure and outcome start and after outcome start? How bad are the cases prognosis?

Features and Functionalities

Defining a target cohort as a set of patients with an exposure or interest and/or with evidence of having an indication of interest and an outcome cohort as a set of patients with evidence of the outcome of interest, we run the following analyses:

1. Cohort Summary - computes aggregate covariate summaries for cohorts (targets and/or outcomes), offering a granular view of the cohort’s demographics, conditions, drug exposures, and more. This enables a deeper understanding of the cohort’s characteristics at various time points at or relative to the index date.:
   • Database Comparison Lets you compare the same cohort across two or more databases and adds in the standardized mean different calculation when exactly two databases are selected. This is a measure of association between the feature and the cohort, therefore identifying which features differ across databases.
     
   • Cohort Comparison Lets you compare two or more cohorts across a database and adds in the standardized mean different calculation when exactly two cohort are selected. This is a measure of association between the feature and the cohort, therefore identifying which features differ across databases.
2. Exposed Case Series - characterizations that look at people in the target cohort who have the outcome during some specified time-at-risk:
   • Risk Factor Compares aggregate covariate summaries for patients in the target who have the outcome during the time-at-risk period vs patients in the target who do not have the outcome during the time-at-risk period. The standardized mean difference is added to identify covariates that differ between the cohorts.
   • Case Series Compares aggregate covariate summaries before target start, between target start up to outcome start and after outcome start for people in the target cohort who have the outcome during some specified time-at-risk. This lets you see covariates that are common before exposure and what happens afterwards.
   • Time to Event: Shows the distribution of when the outcome occurs relative to the target start. This can show you whether the outcome occurs more after or before target exposure.
   • Dechallenge Rechallenge: Offers the ability to compute dechallenge (withdrawal of a drug or treatment) and rechallenge (reintroduction) results. This analysis is critical for understanding the causality between exposures and outcomes, especially in pharmacovigilance studies and when adverse events following exposure to a drug may occur.
3. Incidence Rate: Utilizing the CohortIncidence R package, this set of analyses computes incidence rates for both target and outcome cohorts during the time at risk selected. This feature is essential for assessing the frequency of outcomes or conditions within the specified timeframe, providing a quantitative measure of risk or occurrence. Incidence measures are provided in both tabular and graphical form, and can be stratified across calendar year, age, and sex.

Utility and Application

Characterization serves as a powerful tool for researchers aiming to dissect and understand the nuances of patient cohorts in observational health data. Its capabilities allow for the detailed examination of cohort attributes, the incidence of health outcomes, and the effects of treatment exposures over time. By facilitating a comprehensive analysis of target and comparator cohorts, Characterization enables researchers to draw meaningful conclusions about patient care, treatment efficacy, and health outcomes, thereby contributing to the advancement of evidence-based medicine.

To find out more about the analyses execution details and see examples, please see here.

To see the code behind the Characterization R package, please see here.