How AI might help identify adverse drug events

The development of natural language processing (NLP) capabilities is opening new avenues for pharmacovigilance and pharmacists can play a key role in the process.

US authors have undertaken a narrative review examining the potential for identifying adverse drug events (ADEs) through using artificial intelligence to analyse data from sources such as clinical notes, internet-based data, published medical literature and formally structured data within electronic health records (EHRs).¹ Such analysis has been limited by discrete information ‘silos’, be it by data format or location, which remains in NLP, although work is continuing to reduce this barrier.

While NLP has been studied since the 1980s in relation to detecting medication events,² the amount of electronic and real-time data available now has meant the area has gained increasing attention in the past few years.³

‘The main benefit of NLP is in the time savings associated with automation of various medication safety tasks such as the medication reconciliation process facilitated by computers, as well as the potential for near-real time identification of adverse events for postmarketing surveillance such as those posted on social media that would otherwise go unanalysed¹,’ the authors wrote in their narrative review of the existing research.

Nuances such as the context in which a word was used in case notes (e.g. did the patient have a cold temperature, or the common cold virus?), and the reporting of ‘false positives’ (e.g. ‘he is no longer vomiting’) has presented an issue with the system being able to run unassisted. Instead, a ‘semi-supervised’ approach features a clinician identifying the observations made by the algorithm, and the program subsequently adjusting its learning method. The authors identified semi-supervised and unsupervised methods as an area for further research and development.

Like much of digital clinical data available at the moment, machine learning and NLP is still limited by the amount of information it has access to, and so the issue of interoperability persists.

‘NLP is limited by a lack of data sharing between health care organizations due to insufficient interoperability capabilities, inhibiting large-scale adverse event monitoring across populations,’ the authors wrote.¹

‘In serving as medication experts, pharmacists can provide a key role in serving as an experienced reviewer in supervised machine learning annotation tasks or as a builder of knowledge bases, logistic rules, dictionaries, and other linguistic artifacts for linguistic-focused methods, teaming up with informaticians and computer scientists to improve medication safety.’

‘Continued efforts to improve the utility of clinical NLP include addressing the challenges of characterizing the context of adverse events. Involvement of pharmacists in development of NLP systems will enhance the ability of these systems to improve medication safety.’ ¹

Read the full article here (paywall).

References

1 Wong A, Plasek JM, Montecalvo SP, et al. Natural Language Processing and Its Implications for the Future of Medication Safety: A Narrative Review of Recent Advances and Challenges. Pharmacotherapy. 2018; Jun. doi: 10.1002/phar.2151.

2 Evans RS, Larsen RA, Burke JP, et al. Computer surveillance of hospital-acquired infections and antibiotic use. JAMA 1986;256:1007-11.

3 Luo Y, Thompson WK, Herr TM, et al. Natural language processing for EHR-based pharmacovigilance: a structured review. Drug Safety 2017;40:1075-89.