Could Search Engine Data Improve Consumer Safety?
Studies from Microsoft researchers reveal that search engine data logs may prove a more efficient method of collecting data from pharmaceutical drug users, and could prove to be invaluable assets in the early detection of unexpected side effects.
Over the past year, Microsoft researchers have conducted two separate studies which indicate that search engines may be as effective, or more effective, than the FDA in early detection of unknown drug side effects. These studies demonstrated that search engine log data can be used for the early detection of unknown side effects. They also showed that search engines can be used to predict if a drug recall will occur within the next one to 40 days. The FDA acknowledges that there are some drawbacks to its current drug monitoring program. These studies suggest that search engine data could bolster the FDA’s ability to keep track of drugs’ unexpected side effects,ultimately improving safety conditions for millions.
Problems with the FDA’s Identification of Unknown Side Effects
The FDA’s current program for detecting serious problems with pharmaceuticals is called MedWatch. Its purpose is to gather user reports of “serious reactions, product quality problems, therapeutic inequivalence/failure, and product use errors with human medical products, including drugs, biologic products, medical devices, dietary supplements, infant formula, and cosmetics.” Put simply, MedWatch is a way for consumers to report any and all problems they are experiencing with FDA-regulated health products. But the FDA acknowledges that MedWatch is far from perfect. The program can fall victim to some basic problems associated with self-reporting, such as:
- Unreported Side Effects: Not everyone who experiences a problem with their drug or device will report the issue to the FDA, which slows the organization’s ability to find and address problems with the drug.
- Quality of Information: MedWatch reports contain different levels of detail. Although there are some required fields, such as the name of the product in question, users can share however much – or however little – information they’d like.
- Different Reporting Parties: MedWatch offers self-reporting options, as well as a form for doctors. Different levels of detail and health expertise can make it harder for the FDA to interpret their findings.
Because self-reporting is sporadic, and often incomplete, it takes the FDA a longer period of time to gather evidence of side effects. During that time, thousands of people are put at risk for experiencing these dangerous side effects.Negative drug reactions are the fourth leading cause of death in the United States.
Using Search Engine Data for Early Detection of Unknown Side Effects
In February 2016, a group of Microsoft researchers conducted a study to see whether it was possible to detect unknown drug side effects based on Bing searches. Their study compared the number of FDA reports regarding negative drug reactions to the volume of Bing searches on the same topic. Researchers looked at both the number of times individual users made side-effect-related searches and the overall volume of searches made from one year to six months before the FDA ordered drug label changes.Their research showed that:
- Search engine data is a reliable predictor of previously unknown side effects.
- Search engine data and FDA data are about equally effective in detecting negative side effects.
- Search engine data can identify side effects faster than the data collected by the FDA.
If the FDA implemented search engine detection strategies, unknown side effects might be found sooner and fewer people would have to experience them. While search engine data can still be considered self-reported, it provides a much larger representation of drug users and makes it possible to discover potential prescription drug related problems in real-time.In 2013, eight in 10 online health queries started at a search engine.
Algorithm Uses Search Engine Data to Predict Drug Recalls
Microsoft researcher Elad Yom-Tov trained a machine learning device to predict if a drug would be recalled in the next one to 40 days based on the number of user searches relating to the drug. The device tested over 5,00 drugs and was able to predict recalls with nearly 80% accuracy when predicting a recall one day ahead. Although the algorithm was less accurate at predicting recalls further in advance, its success supports the idea that strategic use of search engine data could improve the FDA’s ability to monitor prescription drugs after they have entered the marketplace.
The Future for Early Detection and Consumer Safety
These studies were not conducted by the FDA, but they may represent a technological turning point for the administration. If search engine proves faster than MedWatch in discovering when drug users are experiencing unexpected side effects, the FDA should take note and consider expanding its methods of drug user data collection.