Published: Systematic Review of Transportability & Generalizability Analysis

We at CCS are proud to announce our the publication of our systematic review of applied transportability and generalizability analyses. This review is a landscape analysis intended to give an overview of all of the published applications of transportability and generalizability analyses using real data. By real data, we mean data that was collected from real people rather than data that was generated for methodological research, such as in simulation study.

What is transportability analysis?

Transportability and generalizability analyses provide a way to transport study results from the population in which the original study was conducted to a different, external population. For example, imagine a clinical trial is conducted to study a new intervention but the trial mostly enrols younger people while outside the trial, most people with the condition are older. Using transportability analysis, we can estimate what the effect of the new intervention would be in an older population. We can also tranport treatment estimates across countries. For instance, we can calculate the effect of a treatment in Germany based on a clinical trial conducted in Canada. There are incredible possibilities for transportability analysis.

Why did we do this review?

Transportability analysis is relatively new. So, implementing these methods can be difficult. That’s why we devleoped {TransportHealth}, an open-source R package to help researchers conduct transportability and generalizability analyses. It also has a companion ShinyApp to make these analyses even more accessible. As part of our development we wanted to make sure we had a full understanding of the potential for these methods, and to ensure that we including the most rigorous approaches in our R package.

Our Process

To complete this review we conducted a systematic search of three health research databases: PubMed, CINAHL, and Embase. Because transportability and generalizability analysis is so new, we also supplemented our search with handsearches of seminal works on this topic. All abstracts and full-text papers were reviewed by two team members to make sure that no eligible work was missed.

In total, we screened more than 1700 abstracts, and did a full-text review of 123 studies.

Our Findings

We were surprised by how many studies we found. In total, our search uncovered 68 published studies of transportability and generalizability analyses using real data. These studies included 83 unique dataset pairs, and 99 distinct analyses.

In general, we found that the number of studies being published and variety methods being used was increasing over time. We also found that these methods are being used in wide range of clinical areas.

There were several innovative applications of transportability analysis that have implications for clinical development and implementation science. As well as methods integrating transportability analysis into meta-analysis.

For full details, check-out the article on the Annals of Epidemiology website.