Internet search history analysis for health diagnosis

Perhaps of limited medico-legal interest absent a ‘duty to rescue’, but interesting nevertheless, is an article by Mark Wilson which notes:

In a study published this week in the Journal of Oncology Practice, Microsoft researchers revealed that they can likely identify many Bing users who have pancreatic cancer—even before their disease has been diagnosed.

Using anonymized historical search data, the team pulled out users whose searches indicated that they had pancreatic cancer. Then, they combed through their search histories for signals of early symptoms. Researchers were able to identify 5%-15% of pancreatic cancer from these early searches alone, with false positive rates as low as 1:100,000—with the ability to distinguish paranoid search queries from earnest self-diagnosis. Given that pancreatic cancer is typically detected late, Microsoft’s methodology could theoretically flag users to get screened sooner and, in turn, increase five-year-survival rates by a small but significant margin.

The primary article by Paparrizos & Ors is entitled Screening for Pancreatic Adenocarcinoma Using Signals From Web Search Logs: Feasibility Study and Results.