anna kemp | ascertaining cases of invasive breast cancer in the 45 and up study: a validation study
DESCRIPTION
Anna Kemp gave an update on her research using the 45 and Up Study data at our 2012 Annual 45 and Up Study Collaborators' Meeting. This meeting is an annual event that offers our research partners, supporters and other interested parties the opportunity to receive a comprehensive update on the 45 and Up Study’s progress and updates on research projects that are using the Study resource. The meeting is also an opportunity for researchers, health decision makers and evaluators to engage and discuss the potential for maximising the Study’s value. For more information, visit www.saxinstitute.org.auTRANSCRIPT
Ascertaining cases of invasive breast cancer in the 45 and Up Study: a validation study.
Kemp A, Preen DB, Saunders C, Holman CDJ, Bulsara M, Rogers K, Roughead EE.
Background
• Routinely-collected and self-reported health data are increasingly used to identify health status and service use
• State-based cancer registries are considered the ‘gold standard’ for identifying breast cancer cases for research purposes o However researchers conducting survey-based studies or
working with other datasets (e.g. hospital data) may need to identify cases without linkage to a cancer registry
o ‘Temporary’ access problems with cancer registrations in New South Wales (NSW)
Background continued
• Larger program of work examining use of prescription hormone therapies for invasive breast cancer in Australian practice (e.g. tamoxifen)
• Planned to identify cases of invasive breast cancer through the NSW Cancer Registry o Power calculations based on cases diagnosed 2003-2009
• We received all the datasets needed for the study except Cancer Registry records for 2009
• Can we ascertain 2009 cases of invasive breast cancer using information from the other available datasets?
Objective
• To determine the accuracy of hospital and outpatient services, prescription claims, and self-report for breast cancer in identifying cases of invasive breast cancer on the NSW Cancer Registry
Data sources
• NSW Cancer Registry (gold standard) o date of diagnosis for all invasive breast cancers in NSW
between 1 July 2004 to 31st December 2008
• 45 and Up Study baseline data o self-reported demographic and clinical information
• NSW Admitted Patients Data Collection o hospital separations for all public and private hospitals
• Pharmaceutical Benefits Scheme o claims for government-subsidised dispensed prescription
medicines • Medicare Benefits Schedule
o claims for government-subsidised outpatient procedures and procedures in private hospitals
Breast cancer ‘flags’
• Hospital diagnosis of invasive breast cancer o ≤ 6 months of diagnosis
• Breast cancer surgery o ≤12 months of diagnosis o mastectomy or lumpectomy
• Prescriptions dispensed o ≤18 months of diagnosis o tamoxifen, toremifene, anastrazole, exemestane, letrozole,
goserelin, trastubumab, lapatinib • Breast radiotherapy
o ≤18 months of diagnosis • Self-reported diagnosis breast cancer and age at diagnosis
o within 12 months of date of diagnosis o analysis restricted to self-reports before January 2006
Analyses
• Cases of invasive breast cancer recorded on the Cancer Registry during the study period were compared with flagged (suspected) cases
• Comparison included o flags from individual datasets (e.g. hospital diagnosis) o clinically meaningful combinations of flags (e.g. hospital
diagnosis and a dispensed medicine for breast cancer) • For each flag/s we calculated:
o positive predictive value (PPV) o sensitivity o Specificity
• We sought flags with high PPV (>90%) and within that, the highest specificity
Results
• Of 143,010 women in the 45 and Up Study, 2661 (1.9%) had a recorded invasive breast cancer on the Cancer Registry during the study period o 681 occurred between 1 July 2004 and 31st December
2005 (this subgroup was compared against self-reported breast cancer)
Results from individual datasets PPV Sensitivity Specificity
45 and Up Study baseline survey
Self-reported diagnosis (12 month window)
40.9% 73.0% 99.5%
Hospital data Inpatient primary diagnosis 80.3% 85.2% 99.7%
Lumpectomy 99.3% 61.3% 99.9%
Mastectomy 99.2% 35.2% 99.9%
Lumpectomy OR mastectomy 99.2% 86.3% 99.9%
Results from individual datasets PPV Sensitivity Specificity
45 and Up Study baseline survey
Self-reported diagnosis (12 month window)
40.9% 73.0% 99.5%
Self-reported diagnosis 72% of the ‘false positives’ had a record on the Cancer Registry for an earlier period
Hospital data Inpatient primary diagnosis 80.3% 85.2% 99.7%
Lumpectomy 99.3% 61.3% 99.9%
Mastectomy 99.2% 35.2% 99.9%
Lumpectomy OR mastectomy 99.2% 86.3% 99.9%
Results from individual datasets PPV Sensitivity Specificity
45 and Up Study baseline survey
Self-reported diagnosis (12 month window)
40.9% 73.0% 99.5%
Self-reported diagnosis 72% of the ‘false positives’ had a record on the Cancer Registry for an earlier period
Hospital data Inpatient primary diagnosis 80.3% 85.2% 99.7%
Lumpectomy 99.3% 61.3% 99.9%
Mastectomy 99.2% 35.2% 99.9%
Lumpectomy OR mastectomy 99.2% 86.3% 99.9%
Results from individual datasets
PPV Sensitivity Specificity
Pharmaceutical Benefits Scheme
Any dispensed medicine 88.5% 68.5% 99.9%
Medicare Benefits Schedule
Breast radiotherapy 95.8% 57.9% 99.9%
Results from 45 and Up Study, MBS and PBS data package
PPV Sensitivity Specificity
Breast radiotherapy AND dispensed medicine
95.8% 41.1% 99.9%
Breast radiotherapy OR dispensed medicine
89.9% 85.3% 99.9%
Breast radiotherapy AND self-reported diagnosis
70.2% 28.3% 99.9%
Breast radiotherapy AND dispensed medicine AND self-reported diagnosis
68.4% 19.4% 99.9%
(Breast radiotherapy OR dispensed medicine) AND self-reported diagnosis
67.8% 56.8% 99.9%
Breast radiotherapy OR dispensed medicine OR self-reported diagnosis
24.9% 94.1% 98.6%
Results from 45 and Up Study, MBS and PBS data package
PPV Sensitivity Specificity
Breast radiotherapy AND dispensed medicine
95.8% 41.1% 99.9%
Breast radiotherapy OR dispensed medicine
89.9% 85.3% 99.9%
Breast radiotherapy AND self-reported diagnosis
70.2% 28.3% 99.9%
Breast radiotherapy AND dispensed medicine AND self-reported diagnosis
68.4% 19.4% 99.9%
(Breast radiotherapy OR dispensed medicine) AND self-reported diagnosis
67.8% 56.8% 99.9%
Breast radiotherapy OR dispensed medicine OR self-reported diagnosis
24.9% 94.1% 98.6%
Results from hospital, 45 and Up Study, MBS and PBS datasets
PPV Sensitivity Specificity
(Lumpectomy or mastectomy) AND diagnosis of invasive breast cancer AND dispensed medicine
99.4% 56.7% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR breast radiotherapy)
96.8% 91.6% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR dispensed medicine)
91.2% 93.7% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR breast radiotherapy OR dispensed medicine)
90.8% 96.8% 99.9%
Results from hospital, 45 and Up Study, MBS and PBS datasets
PPV Sensitivity Specificity
(Lumpectomy or mastectomy) AND diagnosis of invasive breast cancer AND dispensed medicine
99.4% 56.7% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR breast radiotherapy)
96.8% 91.6% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR dispensed medicine)
91.2% 93.7% 99.9%
(Lumpectomy or mastectomy) AND (diagnosis of invasive breast cancer OR breast radiotherapy OR dispensed medicine)
90.8% 96.8% 99.9%
Strengths and weaknesses
• Strengths o Large, heterogeneous sample of women o Complete capture for all public and private inpatient
diagnoses and surgeries, subsidised outpatient procedures and medicines
• Weaknesses o We could not identify how many false positives were DCIS
vs. not breast cancer at all o The flags we have identified may not predict invasive
breast cancer as well in younger women o Validity of the flags examined here may change over time
with shifts in health service use
Conclusions
• Invasive breast cancer can be accurately ascertained through administrative datasets other than the Cancer Registry
• The most useful flags will depend on the research question and available datasets
• Self report with date restriction had moderate sensitivity and low PPV, however specificity was very high o Suitable for excluding cases of breast
cancer from a study sample
Conclusions
• We had access to 45 and Up Study baseline survey, hospital, MBS, and PBS data and needed to identify a sample
• The most useful flag or this purpose: o (Lumpectomy or mastectomy) AND (primary diagnosis of
invasive breast cancer or breast radiotherapy) o 97% of those identified are true positives o 92% of cases on the Cancer Registry
are identified
Acknowledgements
• Participants of the 45 and Up Study • The 45 and Up Study is managed by the Sax Institute in
collaboration with: • Cancer Council New South Wales (major partner) • National Heart Foundation of Australia (NSW Division) • NSW Ministry of Health • beyondblue: the national depression initiative • Ageing, Disability and Home Care NSW Family and
Community Services • Australian Red Cross Blood Service • UnitingCare Ageing
Acknowledgements
• Staff at the Centre for Health Record Linkage • Other data custodians:
o NSW Ministry of Health o Commonwealth Department of Human Services o NSW Cancer Institute
• The project was funded by: o Cancer Australia o National Breast Cancer Foundation
Questions