This Phase II application continues development of the Pharmaceutical Risk Internet Surveillance Monitor (PRISM), an innovative approach to postmarketing surveillance of indicators suggesting diversion/abuse of opioid pharmaceutical agents. Public health relevance of PRISM includes providing reliable and timely data to allow health, law enforcement, industry, and community leaders take steps to limit potential damage associated with an outbreak of pharmaceutical drug abuse. PRISM will be the first, systematically developed and hybrid or partially automated Internet monitoring service to track, at a product-specific level, the chatter of recreational users of pharmaceutical analgesics as they discuss their use/abuse of such drugs. Those who abuse prescription opioids have open access to Web sites, providing an unvarnished picture of users' communications about these drugs, including ideas and beliefs about the drugs as well as discussions of trends and preferences. Thus, Internet monitoring of drug-related messages may be useful for anticipating an increased risk of an abuse outbreak. The potential of these Internet data has not been exploited in a systematic manner to inform drug policy and planning. In Phase I, collaboration with software developers of MIT's General Inquirer (GI) explored the contributions of automated natural language processing methods to the unstructured, fragmented, but revealing messages posted by the online community of substance abusers. Reliable codes were achieved by human coders, supporting the feasibility of classifying content of Internet posts into meaningful categories of endorsing and discouraging abuse of different products. Completely automating such coding may be as yet out of reach. Furthermore, such codes may not have been the most meaningful qualitative information to draw from the posts. However, the overall level of chatter, based on counts of mentions of specific products, did appear to relate to measures of the attractiveness for abuse of target products. There are significant benefits of archiving such posts, using computer processing of raw posts to prep them for human coders, and searching among thousands of archived posts for selected topics of interest (e.g., sources, extraction methods, adverse events, etc.). In Phase II, we will (1) develop software to find, harvest, and archive posts from eligible Web sites for subsequent analyses, (2) refine the nature and types of qualitative questions, (3) develop and test methods of using automated processing of raw posts to improve the reliability and efficiency of human coding, (4) continue exploration of methods to enhance automation of message coding, (5) refine automated data mining of posts discussions on specific topics including (extraction, sources of drug, extreme reactions, etc.) to target drugs, and (6) test external validity of data collected by PRISM. Given the extraordinary financial implications for pharmaceutical companies needing to demonstrate to the FDA that they have an adequate Risk Minimization Action Plan prior to approval for new, potentially addictive medications, a tool like PRISM holds remarkable commercial appeal. ? ? ? ? ? ? ? Public Health Relevance: A system like PRISM to monitor the chatter about abusable substances on Internet is widely recognized as having extraordinary public health importance by the FDA, DEA, and pharmaceutical companies. Ongoing, systematic collection, analysis, and interpretation of abuse trends detected from these Internet sources will be essential to the planning, implementation, and evaluation of public health programs, closely integrated with timely dissemination of these data to those responsible for prevention and control. ? ? ? ?
|McNaughton, Emily C; Coplan, Paul M; Black, Ryan A et al. (2014) Monitoring of internet forums to evaluate reactions to the introduction of reformulated OxyContin to deter abuse. J Med Internet Res 16:e119|
|McNaughton, Emily C; Black, Ryan A; Zulueta, Mirella G et al. (2012) Measuring online endorsement of prescription opioids abuse: an integrative methodology. Pharmacoepidemiol Drug Saf 21:1081-92|