Reasons for failure of an algorithm to perform as expected - analytical:
(1) difference in analytical method and/or reference range for a laboratory test
(2) differences in samples and sample handling
(3) difference in quality and type of imaging studies
Reasons for failure of an algorithm to perform as expected - clinical:
(1) differences in clinical expertise
(2) differences in management, including therapies available
(3) differences in resources available
Reasons for failure of an algorithm to perform as expected - population:
(1) differences in population (age, gender, race)
(2) difference in severity of target disorder
(3) differences in comorbidities
(4) failure to follow inclusion and exclusion criteria
Reasons for failure of an algorithm to perform as expected - analytical:
(1) error in study design or data analysis at initial site (overfitting of data, hidden bias, study size, etc)
(2) difference in time intervals
(3) differences in outcome measures
