TransportHealth: A Novel, Flexible, Open-Source R Package for Transportability and Generalizability Analysis
Author(s)
Quang Vuong, MSc1, Rebecca Metcalfe, BA, MA, PhD2, Richard Yan, MSc3, Jay J. Park, PhD4;
1Core Clinical Sciences, Vancouver, BC, Canada, 2Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 3Simon Fraser University, Department of Statistics and Actuarial Science, Vancouver, BC, Canada, 4McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
1Core Clinical Sciences, Vancouver, BC, Canada, 2Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 3Simon Fraser University, Department of Statistics and Actuarial Science, Vancouver, BC, Canada, 4McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
OBJECTIVES: Transportability and generalizability analyses are causal inference methods that quantitatively assess the external validity of randomized clinical trials and observational studies. In practice, data availability and concomitant methodological requirements for transportability analyses are complex and diverse. The absence of a standardized and validated software that readily performs transportability analysis may limit the uptake of these promising methods. We aimed to develop an open-source R package to perform transportability analysis in the context of varying data availability.
METHODS: Initial steps involved consultation with end-users regarding challenges of conducting transportability analyses in their settings. Following consultation, we began development of statistical functions, as well as supporting materials to meet end-user needs. Comprehensive evaluation was conducted using unit testing and simulation studies to ensure robust performance of the R package.
RESULTS: We developed an R package called TransportHealth to enable broader uptake of transportability analyses. Based on end-user consultation, in addition to the R package, we created supporting vignettes and tutorials to help users understand the nuances of different transportability methods, and how to interpret their results. Due to end-user concerns about the complexity of analysis in R, we developed a companion Shiny application to facilitate analyses by those unfamiliar with the R coding language. To address the diverse use cases for transportability analysis, TransportHealth supports analytic methods that can address varying data limitations, including: inverse odds of participation weighting; g-computation; target aggregate data adjustment; and network meta-interpolation. Unit testing was completed on more than 80% of TransportHealth code. Simulation studies comparing supported methods to naive approaches found that supported approaches yielded lower bias.
CONCLUSIONS: A user-centered approach to package development led to a comprehensive suite of analytic tools for transportability analysis. It remains to be seen if the availability of this open-source, validated R package will increase uptake of transportability analyses.
METHODS: Initial steps involved consultation with end-users regarding challenges of conducting transportability analyses in their settings. Following consultation, we began development of statistical functions, as well as supporting materials to meet end-user needs. Comprehensive evaluation was conducted using unit testing and simulation studies to ensure robust performance of the R package.
RESULTS: We developed an R package called TransportHealth to enable broader uptake of transportability analyses. Based on end-user consultation, in addition to the R package, we created supporting vignettes and tutorials to help users understand the nuances of different transportability methods, and how to interpret their results. Due to end-user concerns about the complexity of analysis in R, we developed a companion Shiny application to facilitate analyses by those unfamiliar with the R coding language. To address the diverse use cases for transportability analysis, TransportHealth supports analytic methods that can address varying data limitations, including: inverse odds of participation weighting; g-computation; target aggregate data adjustment; and network meta-interpolation. Unit testing was completed on more than 80% of TransportHealth code. Simulation studies comparing supported methods to naive approaches found that supported approaches yielded lower bias.
CONCLUSIONS: A user-centered approach to package development led to a comprehensive suite of analytic tools for transportability analysis. It remains to be seen if the availability of this open-source, validated R package will increase uptake of transportability analyses.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR13
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
No Additional Disease & Conditions/Specialized Treatment Areas