Predicting cancer from shopping data
A research team used shopping history records from purchase data tracked on loyalty cards to predict cancer. What should we do with that insight now, though?
In a study of 300 women at Imperial College London, the team found that there were purchasing patterns that may provide some evidence of a change in health. They were able to find links between a diagnosis of ovarian cancer and a history of buying over-the-counter pain and indigestion medications. Early detection of cancer can, of course, have a large impact on health outcomes, but the study raises many ethical issues.
Data availability and privacy laws
Large, population scale data sets, with longitudinal data at a granular level are rare, and loyalty card data is extremely valuable. However, it also subject to a number of biases. There is, for example, often gap between purchase and consumption. A person may buy alcohol or pain medications but have no intention of ingesting it themselves.
The study found evidence of a gap between the women first becoming aware of their symptoms before seeking help from their doctors. Would anyone make the case for sharing purchase data with their GPs? Mandatory sharing would be against privacy laws, but the data is often freely sold to other 3rd parties who do not have the best interests of the individual at heart. There could be a case for voluntary participation in early warning studies.
There is also a little cited clause in GDPR around the duty to use data, where failing to use it may cause harm to individuals. The study raises a number of questions around ethics.
A more appropriate use of the findings in the short term, would be to understand the reluctance of women to consult their doctors in the first instance. If there is a culture of ‘not bothering’ the doctor when important symptoms appear, perhaps it is the culture that needs to change, rather than a widespread snooping policy.
For more on the study, see the original article published in JMIR Public Health and Surveillance.