Validation and application of a death proxy in adult cancer patients

Abstract

PURPOSE: Fact of death is not always available on data sets used for pharmacoepidemiological research. Proxies may be an appropriate substitute in the absence of death data. The purposes of this study were to validate a proxy for death in adult cancer patients and to assess its performance when estimating survival in two cohorts of cancer patients. METHODS: We evaluated 30-, 60-, 90- and 180-day proxies overall and by cancer type using data from 12 394 Australian veterans with lung, colorectal, breast or prostate cancer. The proxy indicated death if the difference between the last dispensing record and the end of the observational period exceeded the proxy cutoff. We then compared actual survival to 90-day proxy estimates in a subset of 4090 veterans with 'full entitlements' for pharmaceutical items and in 3704 Australian women receiving trastuzumab for HER2+ metastatic breast cancer. RESULTS: The 90-day proxy was optimal with an overall sensitivity of 99.3% (95%CI: 98.4-99.7) and a specificity of 97.6% (95%CI: 91.8-99.4). These measures remained high when evaluated by cancer type and spread of disease. The application of the proxy using the most conservative date of death estimate (date of last dispensing) generally underestimated survival, with estimates up to 3 months shorter than survival based on fact of death. CONCLUSIONS: A 90-day death proxy is a robust substitute to identify death in a chronic population when fact of death is not available. The proxy is likely to be valid across a range of chronic diseases as it relies on the presence of 'regular' dispensing records for individual patients. Copyright © 2011 John Wiley & Sons, Ltd.

Studies using Australia's Pharmaceutical Benefits Scheme data for pharmacoepidemiological research: a systematic review of the published literature (1987-2013)

“…30 However, the most recent PBS statistics highlight the growth, particularly in relation to PBS spend, on specialist-initiated medicines such as antineoplastics (cancer drugs). We have very limited research in this important and growing area of therapeutics [31][32][33][34] given cancer represents a significant disease burden in the Australian population. 30 Moreover, populations such as the elderly, children, pregnant women and those with multiple morbidities are studied rarely in pre-market evaluation.…”

Claims-Based Definition of Death in Japanese Claims Database: Validity and Implications

“…Another study in Australia showed that 30 to 180-day death proxy had sensitivity and specificity of 90% or more in adult cancer patients where the proxy indicated death if the difference between the last dispensing record and the end of the observational period exceeded the proxy cutoff [8]. The current study revealed that the combination of the index for ‘dead’ on claims (Definition 1.3) and that for CCIs and 6-month cutoff in the inpatient claim (Definition 2.3) may be used as a composite definition of death (Definition 3.3) to obtain the better sensitivity in researches using Japanese claims data.…”

Does tablet formulation alone improve adherence and persistence: a comparison of ezetimibe fixed dose combination versus ezetimibe separate pill combination?

“…Death data were not available in the PBS dataset. In the absence of this, a sensitivity analysis was undertaken censoring people who ceased all PBS medicines within the 3 months following cessation of the study medicines [39].…”

“…In all analyses, patients were censored when they received their last prescription within 90 days (one prescription supply plus 60 days) from end of follow‐up. Death data were not available in the PBS dataset. In the absence of this, a sensitivity analysis was undertaken censoring people who ceased all PBS medicines within the 3 months following cessation of the study medicines .…”