Monitoring the opioid epidemic via social media discussions

Abstract

The opioid epidemic persists in the U.S., with over 80,000 deaths annually since 2021, primarily driven by synthetic opioids. Responding to this evolving epidemic requires reliable and timely information. One source of data is social media platforms. We assessed the utility of Reddit data for surveillance, covering heroin, prescription, and synthetic drugs. We built a natural language processing pipeline to identify opioid-related content and created a cohort of 1,689,039 Reddit users, each assigned to a state based on their previous Reddit activity. We measured their opioid-related posts over time and compared rates against CDC overdose and NFLIS report rates. To simulate the real-world prediction of synthetic opioid overdose rates, we added near real-time Reddit data to a model relying on CDC mortality data with a typical 6-month reporting lag. Reddit data significantly improved the prediction accuracy of overdose rates. This work suggests that social media can help monitor drug epidemics.

Fig. 1. Growth of opioid discussions amongst all users on Reddit.

a The discussion about opioids on Reddit over time for the top 10 most discussed opioids. Fentanyl has become the most frequently mentioned non-heroin opioid in recent months. b Comment counts for opioids categorized by mu-opioid receptor activity. Mentions of full agonist opioids, those with the highest pain relief and non-medical use potential, have been on the rise since 2010 at a rate similar to heroin mentions and fully surpassing heroin mentions by mid-2017. This change in rates is not identifiable from analyzing the individual opioids.

Fig. 2. National trends in Reddit opioid discussion and CDC opioid overdose rates.

a–c 12-month trailing monthly U.S. opioid overdose CDC death rates per 10,000 people (top panel of subplots, in blue), and monthly opioid comment rates on Reddit per 10,000 total comments (bottom panel of subplots, in orange) for different drug categories, along with d–f: the results of 6-month lead/lag cross-correlation analysis. The cross-correlation analysis indicates that for synthetic opioids, a lead/lag of zero yields the best correlation, indicating that while the magnitude of changes may differ, the shape of the Reddit comment rate over time vs. the benchmark over time is relatively synchronous. The same trend is observed for heroin. Natural and Semi-synthetic CDC trends lag behind Reddit trends by about 5 months. The COVID-19 pandemic is shaded in red from January 2020 to December 2021 and is excluded from the cross-correlation analysis.

Fig. 3. Regional trends in Reddit opioid discussion and NFLIS report rates.

Fentanyl (a), heroin (b), and hydrocodone (c) semi-annual NFLIS opioid report rates per 100,000 people (orange line, left axes) and semi-annual Reddit opioid comment rates per 100,000 total comments (blue line, right axes) from 2014 to 2022 for, plotted at the national (left) and regional levels. We observe similar trends between Reddit and CDC rates for all drugs except during the COVID-19 pandemic, shaded in red from January 2020 to December 2021.

Fig. 4. Autoregressive Integrated Moving Average (ARIMA) simulations for synthetic opioids.

a Methods summary. As an example, to obtain the mortality rate estimate for January 2023, we combined 6-month-lagged CDC data (blue trendline) with 1-month-lagged Reddit data (orange trendline). Grey bars show the observed monthly overdose death rates per 10,000 people as reported by the CDC once available; solid lines represent data incorporated into the prediction, but dashed line data is not included. In this example, the predicted overdose death rate for January 2023 is generated by CDC data from July 2022 alone (left) or in combination with Reddit data from December 2022 (right). b Predicted monthly overdose death rates per 10,000 people of rolling-origin forecast models are shown based on 1-month prediction horizons. Observed mortality is shown in grey, and monthly overdose death rates per 10,000 people as predicted by ARIMA models fitted on 6-month-lagged CDC overdose death (blue) are shown along with predictions from a model that additionally included 1-month-lagged Reddit data (orange). c The absolute errors over time of the monthly overdose deaths per 10,000 people predicted by the lagged CDC model alone (in blue) and the combined CDC/Reddit model (in orange); d shows the corresponding distributions.