Fig. 1

Illustration of how quarter-cohorts were defined and used to construct baseline and updated prediction models. A Individuals were assigned to a particular quarter-cohort based on their date of transplant and followed for 1 year, similar to an “incident-user cohort” design [16]. B (1) The baseline model was developed using data from individuals who were transplanted between 2007 and 2009 (i.e., the quarter-cohorts between 2007 and 2009) and tested on the 2010 quarter 1 (Q1) cohort; (2) the baseline model then was updated using the 2010 Q1 cohort data according to each of the five different update strategies (model U1); (3) model U1 then was tested on the 2010 Q2 cohort; (4) model U1 was updated using the 2010 Q2 cohort data according to each of the five different update strategies (model U2); (5) model U2 was tested on the 2010 Q3 cohort; (6) and (7) subsequent updates occur in the same fashion. A similar procedure was used when the update interval was increased to 2, 4, and 8 quarter-cohorts