Evaluation of linked cancer registry and hospital records of breast cancer

Background: Information on the treatment of women with breast cancer in Aus tralia is generally available only from special surveys. Analysis of routin ely collected datasets may be more timely and cost effective, ii the data a re sufficiently accurate and complete. Objective: To evaluate the accuracy and completeness of data on treatment i n linked records of breast cancer from two routinely collected datasets. Methods: The NSW Department of Health linked NSW Central Cancer Registry (C CR) records for 2,636 women diagnosed with breast cancer in NSW in 1992 to all hospital admission records in the NSW In-patient Statistics Collection (ISC) from January 1991 to June 1994. We queried the original paper records of subsets of women to identify missing or miscoded information and cases not notified to the CCR. We also compared the treatment data with data coll ected independently from the medical records of 19% of the women. Results: ISC records linked to 89% of the CCR records. The CCR had identifi ed 94.9% of women with breast cancer treated as hospital in-patients and 83 % of these women had surgical treatment recorded in the ISC. The linked dat aset under-estimated the percentage of women having breast-conserving thera py (-4%) and slightly over-estimated the percentage having mastectomy (+1%) . We estimated that 42% of women treated surgically for breast cancer had a ctually had breast-conserving surgery, compared with 39% in the original da taset. There was no evident bias by age or by urban or rural residence in t he under-recording of breast conservation. There was 94% agreement on the t ype of surgery between the linked dataset and the independent dataset.