Identifying cases or early breast cancer in Australia using prescription and other health services claims and self-report
Kemp, A., Preen, D., Rogers, K., Saunders, C., Holman, C. D., Bulsara, M., & Roughead, E. E. (2012). Identifying cases or early breast cancer in Australia using prescription and other health services claims and self-report. Pharmacoepidemiology and Drug Safety, 21 (Supplement 3), 8.
Routinely-collected and self-reported health data are increasingly being used to identify health status and service use. Australian state-based cancer registries are the ‘‘gold standard’’ for identifying breast cancer, but researchers working with other datasets (i.e., prescription claims) may need to identify cases without linkage to these registries.
To determine the validity of prescription claims for selective estrogen receptor modulators (SERM) and aromatase inhibitors (AI), hospital procedures, claims for outpatient procedures and radiotherapy, and selfreport in identifying cases of invasive breast cancer in Australia against the Cancer Registry.
Invasive breast cancers recorded on the Cancer Register between 2004 and 2008 for women in the New South Wales 45 and Up Study were compared with cases identified by: (1) SERM and AI prescription claims and (2) outpatient procedures and radiotherapy from 2004 to 2009; (3) NSW Admitted Patients Data Collection (hospital records) between July 2004–February 2009; and selfreported diagnosis of breast cancer between 2003 and 2009 in the 45 and Up Study baseline survey. Sensitivity, specificity, positive and negative predictive values (PPV and NPV) were calculated for each dataset compared with the Cancer Registry.
Of 143,010 women in the 45 and Up cohort, 2,661 (1.9%) had breast cancer recorded on the registry. Sensitivity for self-report of breast cancer diagnosis was 73.0%, with hospital records, PBS and MBS claims data being 86.4%, 65.7% and 58.0%, respectively. PPV was highest for hospital (84.0%) and MBS data (80.4%) and lower for self-report (40.9%) and PBS claims (49.4%). Specificity and NPV were >99% for all comparison datasets evaluated.
In the absence of cancer registrations, cases of breast cancer were most accurately detected using hospital records, and to a lesser extent self-report. Prescription and outpatient claims had only moderate sensitivity and/or PPV, likely reflecting that not all patients have post-surgical pharmacological or medical treatment. However, all of the datasets accurately identified cases without breast cancer, so are suitable for researchers wishing to exclude breast cancer cases from their data.
abstract from conference