To assess the validity of retrospective medical chart review as a method of
classifying prostate-specific antigen (PSA) tests as screening or diagnost
ic services. we reviewed PSA tests ordered at a university hospital (n = 95
). PSA tests were reviewed by four raters: medicine resident (RES), oncolog
ist (ONC), urologist (UR), medicine attending (GM)-and the physician who or
dered the PSA test (ATTEND) using predefined standardized criteria. Agreeme
nt rates by individual rater and ATTEND were 0.79 (GM), 0.80 (ONC), 0.74 (U
R), 0.83 (RES), for a composite percent agreement of 0.79. ATTEND incorrect
ly classified seven tests: exclusion of these tests raised agreement rates
to 0.86 (GM) 0.86 (ONC), 0.80 (UR), 0.90 (RES), for a group composite perce
nt agreement of 0.86. Of note, two raters had higher agreement rates when e
valuating screening PSA tests than when evaluating diagnostic PSA tests. St
andardized criteria applied to medical charts provide a valid method of ret
rospectively classifying PSA tests. (C) 2001 Elsevier Science Inc. All righ
ts reserved.