I was listening to a podcast the other day and was introduced to the idea that google search trends are a very powerful tool in predicting many different things. Some of these have real life implications in medicine. I looked it up and sure enough research has been done to see if specific search terms can predict if a user is developing certain medical conditions. I thought it would be interesting to explore this topic.
In the Journal of Oncology Practice, an article was published looking at the ability of google searches to predict pancreatic cancer. The researchers looked for people who queried the internet for precursor symptoms to pancreatic cancer such as jaundice or itchy skin, light stool, dark urine, sudden weight loss, and abdominal swelling or pressure among other symptoms. They then cross referenced this with users who were eventually diagnosed with pancreatic cancer. Their final results showed “that we can identify 5% to 15% of cases, while preserving extremely low false-positive rates (0.00001 to 0.0001).” In the case of pancreatic cancer this is important because early detection can translate to longer survival rates.
Another study done in the JAMA Dermatology journal sought to determine if internet searches correlate with incidence and mortality rates of common cancers. They came to the conclusion that “population-level internet search behavior may be a valuable real-time tool to estimate cancer incidence and mortality rates” in colon cancer, lung cancer, lymphoma, melanoma and thyroid cancer. This positive correlation is groundbreaking and can have real-world utility. While this study was population based, if we were able to identify these patients through searches and inform them to seek care and screening tests, cancers could be found early and there would be decreased illness burden and death.
Lest we think that data mining can yield only positive results, let’s look at one more article where the findings were not as compelling. In Nature in 2013, an article was published showing google’s flu trend drastically overestimated peak flu levels when compared with the CDC’s real world data. It is a sobering reminder that as “flu-tracking techniques based on mining of web data and on social media proliferate… they will complement, but not substitute for, traditional epidemiological surveillance networks.” While this is an example of where data mining failed, this was in a more public health related capacity and not on the individual level to predict cancer.
So based on the studies above, it does appear that Google has the data and ability to predict if at least some individuals were at an increased risk for cancer. What may be the downside of this though? First, arguments against the practice of data mining should be put aside considering this is already done. We’ll forgo that discussion. There is the worry of false positives and alarming those who may not have cancer, even though this may be rare. In my opinion this would be worth it to save a few lives (and the symptoms those without cancer are googling should probably be seen by a doctor anyway).
What if something was missed? Obviously if people are relying on being told if they cancer or not, there may be repercussions if they are not told. Then there is the worry that people may rely more on google than their doctors. As the Nature study above states, googling should compliment but not substitute the evaluation of a doctor. I think all of these are legitimate concerns.
In conclusion, I think the possibility of having Google predict cancer is an exciting frontier to explore. There are some downsides that need to be considered though. Google and other big name tech companies are already mining our data and using it to make our web experiences more tailored to us an individuals. Should this data be used to improve our health? Do these companies have an ethical duty to warn a person if their search histories are in line with a possible cancer? Are there other downsides? I’d like to know your thoughts. Comment below and add to the conversation.