Abstract
Motivated by limitations of adverse drug reaction (ADR) detection in clinical trials and passive post-market drug safety surveillance systems, a number of researchers have examined social media data as a potential ADR information source. Twitter is a particularly attractive platform because it has a large, diverse user community. Two challenges faced in applying Twitter data are that ADR descriptions are infrequent relative to the overall number of user posts and human review of all posts is impractical. To address these challenges, we framed the ADR detection problem as a binary classification task, where our objective was to develop a computational method that can classify user posts, known as tweets, relative to the presence of an ADR description. We developed a convolutional neural network model (ConvNet) that processes tweets as represented by word vectors created using unsupervised learning on large datasets. The ConvNet model achieved an F1-score of 0.46 and sensitivity of 0.78 for tweet ADR classification on the test dataset, compared to 0.37 F1-score and 0.33 sensitivity obtained by two baseline support vector machine (SVM) models that incorporated word embedding, n-gram, and lexicon features. We attribute the superior ConvNet model performance to its ability to process arbitrary length inputs, which allows it to evaluate every word embedding in a given tweet and make better use of their semantic content as compared to the SVM models which require a fixed length, aggregated embedding input. The results presented demonstrate the feasibility of detection of infrequent ADR mentions in large-scale media data.
Similar content being viewed by others
Notes
We originally sampled 5000 tweets. However, we found that one of our search terms, liquado, a phonetic misspelling of the drug LiquADD, is also a common misspelling of licuado, a blended beverage similar to a smoothie. We therefore eliminated these samples.
A complete description of Lucene scoring is available at http://www.lucenetutorial.com/advanced-topics/scoring.html.
The optimal threshold in the given example was selected using the test set. Results are for illustrative purposes only. As with any model tuning parameter, selection of a threshold that yields generalizable performance would require a cross-validation study.
References
Leaman R, Wojtulewicz L, Sullivan R, Skariah A, Yang J, Gonzalez G (2010) Towards Internet-age pharmacovigilance: extracting adverse drug reactions from user posts to health-related social networks. In: Proc. 2010 Work. Biomed. Nat. Lang. Process., Association for Computational Linguistics, Stroudsburg, PA, USA, pp. 117–125. http://dl.acm.org/citation.cfm?id=1869961.1869976
Benton A, Ungar L, Hill S, Hennessy S, Mao J, Chung A, Leonard CE, Holmes JH (2011) Identifying potential adverse effects using the web: a new approach to medical hypothesis generation. J Biomed Inform 44:989–996. https://doi.org/10.1016/j.jbi.2011.07.005
Yang CC, Jiang L, Yang H, Tang X (2012) Detecting signals of adverse drug reactions from health consumer contributed content in social media. Proc ACM SIGKDD Work Heal Informatics
Yates A, Goharian N (2013) ADRTrace: detecting expected and unexpected adverse drug reactions from user reviews on social media sites. In: Serdyukov P, Braslavski P, Kuznetsov S, Kamps J, Rüger S, Agichtein E, Segalovich I, Yilmaz E (Eds) Adv. Inf. Retr. SE - 92, Springer Berlin Heidelberg, pp 816–819. doi:https://doi.org/10.1007/978-3-642-36973-5_92.
White RW, Tatonetti NP, Shah NH, Altman RB, Horvitz E (2013) Web-scale pharmacovigilance: listening to signals from the crowd. J Am Med Inform Assoc 20:404–408. https://doi.org/10.1136/amiajnl-2012-001482
Freifeld CC, Brownstein JS, Menone CM, Bao W, Filice R, Kass-Hout T, Dasgupta N (2014) Digital drug safety surveillance: monitoring pharmaceutical products in twitter. Drug Saf 37:343–350. https://doi.org/10.1007/s40264-014-0155-x
Ginn R, Pimpalkhute P, Nikfarjam A, Patki A, O’Conner K, Sarker A, Gonzalez G (2014) Mining Twitter for adverse drug reaction mentions: a corpus and classification benchmark. Proc Fourth Work Build Eval Resour Heal Biomed Text Process. http://www.nactem.ac.uk/biotxtm2014/papers/Ginnetal.pdf.
Liu X, Liu J, Chen H (2014) Identifying adverse drug events from health social media: a case study on heart disease discussion forums. In: Zheng X, Zeng D, Chen H, Zhang Y, Xing C Neill DB (eds) Smart Heal. Int. Conf. ICSH 2014, Beijing, China, July 10–11, 2014. Proc., Springer International Publishing, Cham, pp 25–36. doi:https://doi.org/10.1007/978-3-319-08416-9_3.
K. O’Conner, A. Nikfarjam, R. Ginn, P. Pimpalkhute, A. Sarker, K. Smith (2014) Pharmacovigilance on Twitter? Mining tweets for adverse drug reactions. Am Med Informatics Assoc Annu Symp
Nikfarjam A, Sarker A, O’Connor K, Ginn R, Gonzalez G (2015) Pharmacovigilance from social media: mining adverse drug reaction mentions using sequence labeling with word embedding cluster features. J Am Med Informatics Assoc. 22:671–681. https://doi.org/10.1093/jamia/ocu041
Hakkarainen KM, Hedna K, Petzold M, Hägg S (2012) Percentage of patients with preventable adverse drug reactions and preventability of adverse drug reactions—a meta-analysis. PLoS One 7:e33236 10.1371%2Fjournal.pone.0033236
Sultana J, Cutroneo P, Trifirò G (n.d.) Clinical and economic burden of adverse drug reactions. J Pharmacol Pharmacother 73:OP-77 VO-4. doi:https://doi.org/10.4103/0976-500X.120957.
Ahmad SR (2003) Adverse drug event monitoring at the food and drug administration. J Gen Intern Med 18:57–60. https://doi.org/10.1046/j.1525-1497.2003.20130.x
Lindquist M (2008) VigiBase, the WHO Global ICSR Database System: basic facts. Drug Inf J 42:409–419. https://doi.org/10.1177/009286150804200501
Cocos A, Fiks AG, Masino AJ (2017) Deep learning for pharmacovigilance: recurrent neural network architectures for labeling adverse drug reactions in Twitter posts. J Am Med Informatics Assoc 24:813–821. https://doi.org/10.1093/jamia/ocw180
Sarker A, Gonzalez G (2015) Portable automatic text classification for adverse drug reaction detection via multi-corpus training. J Biomed Inform 53:196–207. https://doi.org/10.1016/j.jbi.2014.11.002
Bengio Y, LeCun Y, Henderson D (1994) Globally trained handwritten word recognizer using spatial representation, convolutional neural networks, and hidden Markov models. Adv Neural Inf Process Syst 937–944
Kim Y (2014) Convolutional neural networks for sentence classification. http://arxiv.org/abs/1408.5882 (accessed March 4, 2016)
Kalchbrenner N, Grefenstette E, Blunsom P (2014) A convolutional neural network for modelling sentences. http://arxiv.org/abs/1404.2188
De Boom C, Van Canneyt S, Demeester T, Dhoedt B (2016) Representation learning for very short texts using weighted word embedding aggregation. doi:https://doi.org/10.1016/j.patrec.2016.06.012
Das R, Zaheer M, Dyer C (2015) Gaussian lda for topic models with word embeddings. Proc 53nd Annu Meet Assoc Comput Linguist
Pimpalkhute P, Patki A, Nikfarjam A, Gonzalez G (2014) Phonetic spelling filter for keyword selection in drug mention mining from social media. AMIA Summits Transl Sci Proc 2014:90–95 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4333687/
Abadi M, Agarwal A, Barham P, Brevdo E, Chen Z, Citro C, Corrado GS, Davis A, Dean J, Devin M, Ghemawat S, Goodfellow I, Harp A, Irving G, Isard M, Jia Y, Jozefowicz R, Kaiser L, Kudlur M, Levenberg J, Mane D, Monga R, Moore S, Murray D, Olah C, Schuster M, Shlens J, Steiner B, Sutskever I, Talwar K, Tucker P, Vanhoucke V, Vasudevan V, Viegas F, Vinyals O, Warden P, Wattenberg M, Wicke M, Yu Y, Zheng X (2016) TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems, . http://arxiv.org/abs/1603.04467 (accessed April 8, 2016)
Godin F, Vandersmissen B, De Neve W, Van de Walle Rik (2015) Multimedia lab@ acl w-nut ner shared task: named entity recognition for twitter microposts using distributed word representations, in: ACL-IJCNLP 2015, : pp. 146–153. http://www.aclweb.org/anthology/W/W15/W15-43.pdf#page=158 (accessed April 11, 2017)
T. Mikolov, K. Chen, G. Corrado, J. Dean (2013) Efficient estimation of word representations in vector space. http://arxiv.org/abs/1301.3781
Rehurek R, Sojka P (2010) Software framework for topic modelling with large corpora, in: Proc. Lr. 2010 Work. New Challenges NLP Fram pp. 45--50. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.695.4595 (accessed April 11, 2017).
Nair V, Hinton GE (2010) Rectified linear units improve restricted boltzmann machines, in: Proc. 27th Int. Conf. Mach Learn:807–814
Dahl GE, Sainath TN, Hinton GE (2013) Improving deep neural networks for LVCSR using rectified linear units and dropout. IEEE Int Conf Acoust Speech Signal Process 2013:8609–8613. https://doi.org/10.1109/ICASSP.2013.6639346.
Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R (2014) Dropout: a simple way to prevent neural networks from overfitting. J Mach Learn Res 15:1929–1958 http://jmlr.org/papers/v15/srivastava14a.html
Hinton GE, Srivastava N, Krizhevsky A, Sutskever I, Salakhutdinov RR (2012) Improving neural networks by preventing co-adaptation of feature detectors. http://arxiv.org/abs/1207.0580 (accessed April 28, 2017)
Dietterich TG (1998) Approximate statistical tests for comparing supervised classification learning algorithms. Neural Comput 10:1895–1923. https://doi.org/10.1162/089976698300017197
Johnson AEW, Pollard TJ, Shen L, Lehman LH, Feng M, Ghassemi M, Moody B, Szolovits P, Anthony Celi L, Mark RG (2016) MIMIC-III, a freely accessible critical care database. Sci Data 3:160035. https://doi.org/10.1038/sdata.2016.35
Pan SJ, Yang Q (2010) A survey on transfer learning. IEEE Trans Knowl Data Eng 22:1345–1359. https://doi.org/10.1109/TKDE.2009.191
Funding
The Leonard David Institute at the University of Pennsylvania supported this work.
Author information
Authors and Affiliations
Contributions
AJM and AGF conceived of and designed the study. AJM, DF, and AGF conducted data labeling and review. AJM and DF developed the model. AM and DF implemented the models. AJM and DF analyzed model performance. AJM wrote the manuscript. All authors contributed to the review and revisions of the manuscript. All the authors have seen and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Conflict of Interests
AGF received an independent research grant from Pfizer, which provided salary support for his research team for work unrelated to this project.
Rights and permissions
About this article
Cite this article
Masino, A.J., Forsyth, D. & Fiks, A.G. Detecting Adverse Drug Reactions on Twitter with Convolutional Neural Networks and Word Embedding Features. J Healthc Inform Res 2, 25–43 (2018). https://doi.org/10.1007/s41666-018-0018-9
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s41666-018-0018-9