Systematic review of applied transportability and generalizability analyses: A landscape analysis

Transportability and generalizability analysis are novel causal inference methods that quantitatively assess external validity. Currently, it is unclear how these analyses are applied in practice. To characterize applications and methods, we conducted a landscape analysis of applied transportability and generalizability analyses using a systematic literature search of PubMed, CINAHL and Embase supplemented with hand-searches. We identified 68 publications describing transportability and generalizability analyses conducted with 83 unique source-target dataset pairs and reporting 99 distinct analyses. The majority of source and target datasets were collected in the US (n = 63/83, 75.9 %; and n = 59/83, 71.1 %, respectively). These methods were most often applied to transport RCT findings to observational studies (n = 38/83; 45.8 %), or to another RCT (n = 20/83; 24.1 %). Several studies used transportability analysis outside the standard application, for example to identify effect modifiers or calibrate measurements within an RCT. Methods that used weights and individual-level patient data were most common (n = 56/99, 56.5 %; n = 80/83, 96.4 %, respectively). Reporting quality varied across studies. Transportability analysis has a wide range of applications including supporting decision-making by improving evidence relevance and improving trial design by identifying contextual effect modifiers and calibrating outcome measurements. Efforts are needed to standardize analysis and reporting of these methods to improve transparency and uptake.