The performance of marginal structural models for estimating risk differences and relative risks using weighted univariate generalized linear models
Austin PC. Stat Methods Med Res. 2024; Apr 24 [Epub ahead of print].
Objective — The quality of coding for breast surgical procedures was examined by comparing hospital discharge abstracts and physician claims with data abstracted from records of women diagnosed with node-negative breast cancer from April 1, 1991, to December 31, 1991.
Methods — The node-negative breast cancer cohort was linked with a population registry file. Hospital discharge abstracts and physician billing claims were retrieved for matched subjects. Overall agreement between two data sets was defined as the number of cases for which there was a match by specific type of procedure out of all eligible cases that were matched with the healthcare utilization file. Specific agreement was assessed by the kappa statistic, using only those records in the administrative data set that were coded for mastectomy or breast-conserving surgery.
Results — Of 735 eligible cases in the node-negative breast cancer cohort, 655 (89.1%) were linked to a healthcare utilization file. Overall agreement between surgeon billing claims and charts was 95.4% (CI = 93.5, 96.9) for most definitive procedure. Agreement for breast surgery type was 98.1% (kappa = 0.96; CI = 0.87,1.0) for cases coded as breast-conserving surgery or mastectomy. When hospital discharge and chart data were compared, overall agreement was 86.2% (CI = 83.4, 88.8), whereas agreement for breast surgery type was 93.2% (kappa = 0.86; CI = 0.77, 0.94).
Conclusion — Overall, definitive surgical procedure in the two administrative databases accurately reflected information recorded in patients' charts. Physician claims appeared to provide more accurate information than did hospital discharge data.
Pinfold SP, Goel V, Sawka C. Med Care. 2000; 38(1):99-107.
The ICES website uses cookies. If that’s okay with you, keep on browsing, or learn more about our Privacy Policy.