The United States is in the midst of an opioid crisis. Every day, over 100 people die from opioid overdoses, according to the US Centers for Disease Control and Prevention. But what if we could know about overdoses before they happen?
Scientists in California have opened the possibility of having such preemptive knowledge by creating a model that uses Google searches to predict overdoses from heroin. Their research, published in the journal Drug and Alcohol Dependence in September and reported on this week in Scientific American, shows that Google searches for certain drugs, including slang terms, can be used to explain heroin-related visits to hospitals.
The researchers used data from nine US metropolitan areas on Google searches for opioids (both prescription and not), and data on heroin-related visits to the emergency department, from 2004 to 2011. The model used search terms one year to predict emergency-room visits the next. It included the keywords “Avinza,” “Brown Sugar,” “China White,” “Codeine,” “Kadian,” “Methadone,” and “Oxymorphone,” and could explain 72% of the variation in heroin-related hospital visits. The authors found that a higher number of searches per keyword was associated with more overdoses.
There’s a long way between making the initial finding and putting it to practical use. As a comparison, scientists have created models that are extremely accurate at predicting whether someone is at risk of suicide, but it will take years and a trial in a hospital before such results can be put to widespread use.
And the model needs some work: To be effective outside the study, it should be able to distinguish “brown sugar” the sweetener from the slang term for heroin. Still, as the US is faced with a seemingly unending opioid crisis, the ability to predict heroin deaths provides the hope that there are still new means of tackling the epidemic.