By Shiya John, Sri Krishna Kaashyap Madabhushi, Nishitha Chidipothu, Jim Samuel
The opioid crisis is affecting all levels of American society regardless of class, ethnicity, gender, or career and there is an urgent and compelling need to address all stages of the opioid crisis lifecycle. This dilemma must be viewed as a dynamic, ever-evolving problem that has psychological, emotional, and biological implications that can impact many vulnerable segments of society. We define the “opioid crisis lifecycle” (OCL) as consisting of stages which lead to opioid addiction and overdose. OCL involves several stages, including vulnerability, tolerance, dependence, abuse, addiction, and overdose and extreme crisis events. When a person begins to feel an increasing urge to use opioids due to factors other than medical reasons, such as emotional, social, or psychological needs and surroundings, it marks an increase in vulnerability. When a person’s brain becomes accustomed to an opioid, routine dosage loses its effectiveness and “tolerance” occurs. To achieve the same level of pain relief and relaxation, the user then must take higher doses. The body will gradually adapt to regular opioid use and become dependent on the medicines or drugs for normal function. Even when taken as directed by medical practitioners, opioids can cause dependence in anyone who uses them frequently. Abuse includes using drugs that have not been approved by a doctor, using excessive amounts of medication, or receiving prescriptions from multiple physicians. Contrary to vulnerability, tolerance, and dependence, addiction and opioid use disorder refer to medical conditions (1). Any sustainable approach to tackling this crisis must address all stages of the OCL, and in this research-based article, we discuss ways in which informatics and artificial intelligence can help address the opioid crisis.
Excessive use of opioids, whether illicit or prescribed, has severe and often irreversible effects on the mind and body such as depressions, disability, relapses, overdose, and death. Tackling the opioid crisis comprehensively is essential in each of the aforementioned stages. An important line of defence against the opioid problem is the prevention of overdose deaths following abuse and addiction. It is possible to make this happen using informatics, robotics, and AI technologies for prevention, identification, treatment, and rehabilitation. In a study covering 81 publications, including more than 5.3 million participants and 14.6 million social media posts, AI methods were applied to examine opioid use (2). AI is defined as being a “… a set of technologies that mimic the functions and expressions of human intelligence, specifically cognition and logic” and informatics as being “… advanced technology-driven big data analytics” (3)(4). Informatics and AI can be used for vulnerability identification and mitigation, risk prediction, opioid surveillance, scenario monitoring, patient support, and pain management in various stages of the OCL.
There is a tremendous need for investment into AI and Machine Learning (ML), and informatics systems which will give us information about the logistics and process of shipment, distribution, and illegal prescription or drug diversion incidents. Algorithmic ML-driven computer models can learn “to determine the timing of intervention, inhibit drug craving and relapse, detect opioid intake, and classify pre-and post-opioid health conditions” when hospitals, governments, and individuals input more data (5). Developing AI tools that can identify and adaptively update the factors causing opioid addiction may help reduce the probability of having a negative opioid-related outcome. This can be achieved in two ways: either by detecting dangers beforehand and averting the ensuing problem, or by using a system of machine learning-based classification mechanism for dependent and long-term opioid users (2). There has been an increasing number of ML algorithms and AI solutions for opioid overdose risk prediction for Medicare recipients using opioid prescriptions. While analysing the benefits and risks trade-offs of prescription opioid use for high-risk patients, algorithms crafted by researchers’ from the University of Florida can accurately identify 70% to 90% of high-risk patients, excluding the vast majority of users with negligible risk of opioid use disorder or overdose (6). By using mobile apps and individual social media posts, researchers can study the negative health effects of opioid usage, the switch to injecting drugs, suicidality, and opioid overdose among opioid users. Social media data was widely used to study individual and group behaviour during the COVID-19 pandemic (7)(8)(9)(10)(11). Researchers have also identified various subgroups, estimated occurrence, and the scope and location of unauthorized online opioid sales portals and mechanisms using AI (12).
Opioids-based pain management strategies for patients in acute pain can increase the risk of opioid addiction. Patients with recent surgeries fall into this category of high risk for opioid addiction because they may have to use a higher dose of an opioid for a longer period of time (13). The independent predictive variables are “age, higher body mass index, female gender, type of surgery, pre-existing pain, and prior opioid use” (14). Identifying patients with high risk for post-surgical pain can motivate doctors to apply limited-opioids or no-opioids-based pain management strategies. ML algorithms can be used with health records and behavioural data to identify patients at a high risk of post-surgical pain more effectively. Research presented at the Anesthesiology 2020 Annual Meeting, held by the American Society of Anesthesiologists, suggested that ML models using logistic regression, random forest, and artificial neural networks algorithms can accurately identify patients with high post-surgery pain management needs about 80% of the time (14). Logistic regression and random forest models achieved an accuracy of 81% and the artificial neural network model provided an accuracy of 80% (13). Such models will score patients for risks of post-surgery pain and share that information with their anesthesiologists and surgeons, aiding the medical practitioners to develop alternatives to opioids for pain management for those with higher vulnerability (14).
AI has powerful capabilities, that far surpass human intelligence capabilities, to identify patterns and generate insights that provide in-depth and adaptive views for dynamic problems. Agencies will need the technological prowess of AI to reduce the risks associated with the OCL issues that lead to opioid addiction and overdose. The Nassau County Police Department in New York state uses a “real-time overdose mapping technology” which provides timely warnings on overdose patterns and helps save lives (15). Digital health technologies like mHealth apps are invaluable in dealing with opioid use disorder. These apps can help monitor patients to provide personalized care by monitoring their vitals’ trends to predict excessive opioid usage (7). For example, Indiana-based Hc1.com launched their Opioid Dashboard with ML that uses data based on opioid prescriptions and identifies potential patterns of misuse and abuse among individuals who use opioids. Large volumes of data pulled from “3.8 million providers, 51 million individuals and 5 billion diagnostic test results” were used to train their ML models and systems (16). Triggr Health, a Chicago-based company, uses ML to “predict addiction recovery relapses to help target prevention strategies” (16). Their algorithms employ smartphone data which includes texting behaviour, mobile device locations, and sleep patterns among others to estimate the likelihood of a relapse. It is also possible to use wearable biosensors to track and report physiological changes linked to opioid intake or drug cravings, and provide timely individualized preventive interventions (7).
AI applications and informatics used to support the management of the OCL are primarily concerned with identification of patterns, development of behavioural models, prediction, and prevention of excessive or unwarranted opioid usage by processing human-behaviour data and other relevant data associated with the OCL (16)(17). Despite the improved performance of AI applications for addressing the OCL, numerous challenges related to ethical and security concerns linger, including data protection, privacy of sensitive health records, and cybersecurity. Another challenge is for businesses supporting the OCL to build profitable ROI models to ensure sustainability. For-profit entities depending on payments from patients or their relatives could face challenges, considering that “the highest rates of opioid overdoses have occurred in the poorer regions of the country,” and this should motivate companies to look at low cost and effective AI-driven solutions (16). Hence, it is very important for businesses to develop partnerships with government bodies to build sustainable business models and help prevent the opioid crisis from increasing to pandemic-like proportions.
Our review of large quantities of news articles’ headlines from across the world reveal that though the use of AI applications is on the rise, most efforts are fragmented and integration of diverse AI applications across the OCL remains an open problem. If correctly applied, adaptive AI applications can improve the performance of healthcare professionals taking care of patients at various stages of the OCL (18)(19). It is therefore imperative that we, as a society with an emphasized responsibility on government agencies, recognize and take immediate steps towards meeting the need for an artificial intelligence and informatics-based strategy to address the opioid crisis.
- Sreenivas, S. (n.d.). Stages of Opioid Addiction. Retrieved September 10, 2022, from https://www.webmd.com/mental-health/addiction/stages-of-opioid-addiction
- Gadhia, S., Richards, G. C., Marriott, T., & Rose, J. (2022). Artificial intelligence and opioid use: a narrative review. medRxiv.
- Samuel, J. (2021). A call for proactive policies for informatics and artificial intelligence technologies. Scholars Strategy Network. Url: https://scholars.org/contribution/call-proactive-policies-informatics-and
- Samuel, J. (2022). 2050: Will Artificial Intelligence Dominate Humans? (AI Slides). SSRN https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4155700 , June 2022.
- Boltz Director, T. (n.d.). Opioid management: Preventative technology through AI and machine learning. Carahsoft. Retrieved September 18, 2022, from https://www.carahsoft.com/community/carahsoft-opioid-prevention-ai-ml-blog-2022
- Lo-Ciganic, W. H., Huang, J. L., Zhang, H. H., Weiss, J. C., Wu, Y., Kwoh, C. K., … & Gellad, W. F. (2019). Evaluation of machine-learning algorithms for predicting opioid overdose risk among medicare beneficiaries with opioid prescriptions. JAMA network open, 2(3), e190968-e190968.
- Nuamah, J., Mehta, R., & Sasangohar, F. (2020). Technologies for Opioid Use Disorder Management: Mobile App Search and Scoping Review. JMIR mHealth and uHealth, 8(6), e15752. https://doi.org/10.2196/15752
- Ali, G. M. N., Rahman, M. M., Hossain, M. A., Rahman, M. S., Paul, K. C., Thill, J. C., & Samuel, J. (2021, August). Public perceptions of COVID-19 vaccines: Policy implications from US spatiotemporal sentiment analytics. In Healthcare (Vol. 9, No. 9, p. 1110). MDPI.
- Rahman, M. M., Ali, G. M. N., Li, X. J., Samuel, J., Paul, K. C., Chong, P. H., & Yakubov, M. (2021). Socioeconomic factors analysis for COVID-19 US reopening sentiment with Twitter and census data. Heliyon (ScienceDirect by Elsevier), e06200.
- Samuel, J., Rahman, M., Ali, Nawaz G. G. Md., Samuel, Y., Pelaez, A., Chong, P. H. J. and Yakubov, M (2020) “Feeling Positive About Reopening? New Normal Scenarios From COVID-19 US Reopen Sentiment Analytics,” in IEEE Access, vol. 8, pp. 142173-142190, 2020, doi: 10.1109/ACCESS.2020.3013933. https://ieeexplore.ieee.org/document/9154672
- Samuel, J., Ali, G. G., Rahman, M., Esawi, E., & Samuel, Y. (2020). Covid-19 public sentiment insights and machine learning for tweets classification. Information, 11(6), 314.
- Sarker, A., Gonzalez-Hernandez, G., & Perrone, J. (2019). Towards automating location-specific opioid toxicosurveillance from Twitter via data science methods. Studies in health technology and informatics, 264, 333.
- Artificial intelligence can predict patients at highest risk for severe pain, increased opioid use after surgery. (n.d.). Retrieved September 18, 2022, from https://www.asahq.org/about-asa/newsroom/news-releases/2020/10/artificial-intelligence-can-predict-patients-at-highest-risk-for-severe-pain-increased-opioid-use-after-surgery
- Clemens, D. (2020, October 9). Ai machine learning could help tailor opioid treatment, reduce addiction risk. Anesthesia Experts. Retrieved September 18, 2022, from http://anesthesiaexperts.com/uncategorized/ai-machine-learning-tailor-opioid-treatment-reduce-addiction-risk/
- Industries, I. B. M. (2018, April 3). Can ai help stop the opioid crisis? Medium. Retrieved September 18, 2022, from https://medium.com/@IBMindustries/can-ai-help-stop-the-opioid-crisis-194aabf8d06
- Sennaar, K. (2019, May 19). Artificial Intelligence and the opioid epidemic – applications for relapse treatment, abuse prevention, and more. Emerj Artificial Intelligence Research. Retrieved September 18, 2022, from https://emerj.com/ai-sector-overviews/artificial-intelligence-opioid-epidemic/
- McCaney, K. (n.d.). AI algorithms can help combat the opioid crisis. governmentciomedia.com. Retrieved September 18, 2022, from https://governmentciomedia.com/ai-algorithms-can-help-combat-opioid-crisis
- Lin, E. J., Schroeder, M., Huang, Y., & Linwood, S. L. (2022). Digital Health for the Opioid Crisis: A Historical Analysis of NIH Funding from 2013 to 2017. Exon Publications, 33-43.
- Samuel, J., Kashyap, Yana Samuel, and Alexander Pelaez. Adaptive cognitive ﬁt: Artiﬁcial intelligence augmented management of information facets and representations. International Journal of Information Management 65 (2022) 102505, https://doi.org/10.1016/j.ijinfomgt.2022.102505