Accomplishments and Opportunities in Biosurveillance

S urveillance is the cornerstone of public health. Robust surveillance data and their analysis meaningfully impact public health action, planning, and prioritization. However, as Dr Alexander Langmuir stated in his 1962 Cutter Lecture on Preventive Medicine at the Harvard School of Public Health, “Good surveillance does not necessarily ensure making the right decisions, but it reduces the chances of wrong ones.” Over the past 50 years, the application of computer and information science has improved efficiency and effectiveness of public health surveillance by changing the way we collect, process, and analyze vast and disparate data for research, decision making, and learning. One example is the development of syndromic surveillance systems. What began as an experiment in a few large cities 20 years ago to track over-the-counter sales of antidiarrhea medications as an early warning of outbreaks of gastrointestinal illness expanded substantially after the terrorist attacks of 2001. Today the model has expanded to include many additional sources of real-time or near real-time data and broader uses of those data. Health departments are using data collected from emergency department visits that they receive from hospitals for monitoring conditions for which there are no surveillance systems (eg, opioid overdoses and carbon monoxide poisonings) and for improving public health situational awareness. These syndromic-based systems can continue to improve as additional text fields from the electronic health record (EHR) become part of the data stream. These additional data along with advanced natural-language processing and statistical learning methods may enhance the use of other coded and free-text contextual information. Over time, adaptive machine learning methods could make possible the detection of syndromes that were not prespecified, which could enhance overall surveillance and improve early event detection.