Objective: To establish the reliability of a coding system for screening and diagnostic prostate-specific antigen (PSA) testing from patient charts.
Study design and setting: Two investigators reviewed 448 chart abstractions from a population-based case-control study of PSA screening in the Toronto area. The tests evaluated for reliability were transrectal ultrasound (TRUS), digital rectal examination (DRE), and prostate-specific antigen (PSA).
Results: DRE results were found in 87%, PSA results in 65%, and TRUS results in 12% of the 749 charts. Interobserver agreement was 94% for DRE texture (kappa =.885), 95% for DRE asymmetry (kappa = .868), 85% for DRE physician interpretation (kappa = .698), 97% for final DRE result (kappa = .856), and 87% for TRUS (kappa = .769). Physician interpretation modified the final result in only 6.2% of DREs. Interobserver agreement for PSA coding was 91% (kappa = .787). Of PSA results, pure PSA screening with no symptoms of obstructive urination was found in 19%, symptomatic PSA screening in 46%, and diagnostic PSA testing in 35%.
Conclusion: We have developed a practical and reliable coding system for TRUS, DRE, and PSA in the context of a case-control study of PSA screening.