US Pharm. 2018;43(3):21-24.
Information technology continues to play a major role in the diagnosis and delivery of information to stakeholders in the healthcare sector. This article will focus on information technology and infectious diseases from a public-health perspective. Infectious diseases account for a large percentage of healthcare expenditures, significantly impacting the well-being of individuals and communities. For example, 2011 saw well over 3.9 million hospital outpatient-department visits for infectious and parasitic diseases, and in 2014, 17.8 million visits to physicians’ offices for infectious and parasitic diseases were documented.1,2 Morbidity was documented from 9,421 new tuberculosis cases (2014); 51,455 new Salmonella cases (2014); 33,461 new Lyme disease cases (2014); and 433 new meningococcal disease cases (2014).3
Having an ability to rapidly track, analyze, and diagnose various infectious processes in real time using robust data sources not only may assist in creating better mechanisms for pattern tracking and disease prevention, but might also be instrumental in preventing untimely deaths.
Better Decision Making Through Modeling
The practice of monitoring infectious-disease processes has traditionally relied heavily on surveillance and expert opinion. Once the surveillance data are collected, public-health officials consult with subject-matter experts, and appropriate measures to control an infectious-disease outbreak are designed and implemented. However, these actions are not always efficiently coordinated and do not occur quickly enough to enable rapid decision making that can minimize morbidity and mortality. Modeling is a tool that fills the void in preemptive infectious-disease decision making by using available data to provide quantitative estimates of outbreak trajectories. Public-health officials create these models in an effort to intervene early and prevent the spread of an infectious-disease outbreak.
An example of such a tool includes the Susceptible-Infectious-Recovered (SIR) model. This type of mathematical model categorizes individuals according to their status with respect to the infection: susceptible to the infection, currently infectious, or recovered and immune or deceased. Transitions between these states within the mathematical model are controlled by a set of parameters that need to be estimated. The effects of potential measures to control the outbreak can be incorporated into the model to predict its effectiveness. The model can be modified to allow a range of parameter values, rather than a single value for each parameter. This allows public-health officials to perform a sensitivity analysis, which demonstrates how the outcomes could change if there is an error in the estimated values. There are now online applications of the SIR model to facilitate broad adoption.
Modeling is an attractive public-health tool because of its ability to estimate an outbreak’s trajectory and the effects of possible control measures in a timely manner.4
Another model, called ResistanceOpen, compiles online data that are publicly available from various institutions, including community healthcare providers and regional, national, and international organizations, to monitor the emergence of antibiotic resistance. The model displays the information on a map that can be used by public-health officials.5
Trajectory of Artificial Intelligence
While modeling is an improvement over standard documentation, artificial intelligence (AI) technology is evolving at a very rapid rate. The ability of AI technology to augment decision-making processes is attributed to the speed of pattern recognition and the robust amount of data that are digested and analyzed for optimal health outcomes.
There is currently an onslaught of data derived from electronic health records, health exchanges, and other digital sources, which provides real-time, detailed information on various infectious-disease processes and outbreaks. Prior to this, information was derived from clinical laboratories and data collected by the public- health sector; however, this often created an excessive lag time, was difficult to reproduce, and was costly. With the increase in the volume of patient healthcare data available from various electronic and mobile sources, as well as unstructured data such as photos, physician notes, sensor data, and genomic information, it has become more difficult to organize these incoming data streams coherently using traditional computational algorithms.
As a consequence, these data are often not used and reside on servers and in data clouds, with payers, clinicians, and patients unable to realize their potential. With technologies supporting AI, however, these data can easily be captured, catalogued, reviewed, analyzed, and acted upon in an efficient manner. It is hoped that AI will synergize with efforts already in public-health practice that will assist in creating novel approaches to disease surveillance, resulting in prompt disease detection.
AI holds the promise of capitalizing on varied and divergent incoming data streams and makes “intelligent” inferences based on the vast amounts of raw data accumulated. This capability has the potential to produce effective tools to assist in infectious-disease prevention and the discovery of cures. If used effectively and judiciously in the coming months and years, AI’s unparalleled ability to maximize and create new opportunities to detect and monitor infectious diseases will be realized.
Having access to data that are accumulated electronically in real time enables AI information technology to quickly track and trend disease outbreaks, allowing public-health professionals and front-line clinicians to better manage and contain the widespread dissemination of an infectious process. Optimally, technology tools that employ surveillance, together with data compilation and powerful AI analytics, will have the most success in creating a synergy with—rather than simply supplanting—existing health department protocols.5
Exploiting Electronic Health Record Data
At present, simple steps such as utilizing electronic health-record data—with patient identifiers removed—is a potential helpful resource to monitor infectious disease outcomes, vaccine uptake, and adverse drug reactions. These technologies are now widely deployed throughout the healthcare sector, and they provide an opportunity to facilitate change in the way infectious diseases are detected, and potentially diagnosed and treated. AI provides additional information using the most up-to-date research from around the world and includes outcomes figures, allowing immediate responses or modifications to an emerging infectious-disease outbreak. This analysis can occur quickly and efficiently and can also assist in warning of any emerging infections not currently known to exist in a geographic region. The technologies are now being embedded in portable devices, such as cellular phones, for easy access to clinical data at any time in the healthcare process. AI data that are cloud-based can be sent to e-mail listservs, electronic drug databases, and search engines that allow focused clinical questions and answers in real time, with accuracy based on numerous pieces of data accumulated, dissected, and stored for immediate use.
One caveat is that at times, regardless of current electronic tools in place today, adoption of such techniques has proven to be slow in terms of user acceptance. Data sharing, which is critical for the success of the envisioned public- health surveillance system, often does not occur because clinicians err on the side of patient privacy and security concerns. Nevertheless, extensive computer-based software programs have been developed that are essential for physicians and microbiologists.6 Adequate protections can be built into software to protect individual patient privacy while allowing for aggregate data collection.
Analyzing the Human Immune System
Major vendors, such as Microsoft, using AI are partnering with other vendors (e.g., Adaptive Biotechnologies) to decode the human immune system with the goal of leveraging the company’s “machine learning and cloud computing capabilities” to assist in the bioinformatics analysis of DNA sequencing data of T-cell and B-cell receptors, which make up the immune system. Once data are accumulated, AI processes the information to develop a “universal T-cell receptor/antigen map” to detect disease. The ultimate goal in such a case is the creation of a universal blood test that reads a person’s immune system to detect a wide variety of diseases, including infections, cancers, and autoimmune disorders, in their earliest stage.7 The blood test would not only enable researchers to determine an individual’s exposure to a particular disease, but would also help personalize treatments based on the person’s immunological history, including diseases he or she has overcome in the past. This exemplifies AI’s capability for supplementing the human factor and current technologies at its finest.
Another example is IBM’s use of AI supporting healthcare professionals and various other healthcare stakeholders to improve the care and health outcomes of who have or are susceptible to infectious diseases. IBM’s Watson supercomputer software has the ability to unlock insights using a vast array of data. This, in turn, permits prompt actions to manage the disease with best practices as determined by AI, thereby complementing the clinician. In addition, AI solutions provide data integration and aggregation, creating prevention strategies for high-risk patients as well as for the general population.
Financial Investment Is Key
AI provides the technical platform to optimize the current practice of public-health infectious disease surveillance. However, the success of this technology depends on widespread implementation, which will require significant financial investments in AI-enabled devices. State and local public-health departments will require compelling levels of proof that the AI technologies will improve surveillance for infectious diseases, save lives, improve community health, and lower long-term healthcare costs. State and local information technology budgets are already strained with competing priorities, and high levels of proof will be required to convince agencies to commit resources to AI technologies.8
With all of the promise and opportunity that come with using AI in healthcare, due caution must be exercised because AI is only as good as the data collected and dissected. Algorithms utilized by AI need to be based on accurate and unbiased data in order to better inform the healthcare decision-making process. Only then will AI be best positioned to predict future clinical results. Therefore, the most accurate way to deal with AI-supported solutions is to view the system as another synergistic piece of information technology infrastructure that needs to easily accommodate and update components of new data.
Additionally, a data integrity strategy must be in place that can ensure not only practical and timely data, but also proper aggregation and normalization of those data that ultimately supports the building, testing, and deployment of machine-learning algorithms across an organization. This approach will enable organizations to take advantage of a number of industry innovations while reducing the risk that a machine-learning system becomes obsolete and results in costly custom integrations.9
Conclusion
AI systems can be extremely powerful and, if used correctly, very supportive of current healthcare and its growth. However, one element that is often not emphasized enough is the role played by data in any successful AI-supported system. Datasets will prove invaluable as they are collected from the various electronic and social media feeds. Additionally, electronic systems have collected and analyzed numerous publications, all of which are combined to create a diverse database of information for us to learn from and inform healthcare decisions. Thus, AI will assist healthcare providers in using the collected data in more cohesive ways, providing the key ingredients to making smarter health systems that help healthcare providers care for the patient and change the trajectory of healthcare.
REFERENCES
1. Centers for Disease Control and Prevention. National Hospital Ambulatory Medical Care Survey: 2011 outpatient department summary tables. www.cdc.gov/nchs/data/ahcd/nhamcs_outpatient/2011_opd_web_tables.pdf. Accessed January 28, 2018.2. Centers for Disease Control and Prevention. National Ambulatory Medical Care Survey: 2014 state and national summary tables. www.cdc.gov/nchs/data/ahcd/namcs_summary/2014_namcs_web_tables.pdf. Accessed January 28, 2018.3. Centers for Disease Control and Prevention. National Center for Health Statistics. Health, United States, 2016. www.cdc.gov/nchs/data/hus/hus16.pdf#033. Accessed January 28, 2018.4. Wearing HJ, Rohani P, Keeling MJ. Appropriate models for the management of infectious diseases. PLoS Med. 2005;2:Epub.5. Bansal S, Chowell G, Simnonsen L, et al. Big data for infectious disease surveillance and modeling. J Infect Dis. 2016;214(suppl 4):S375-S379.6. Magos A, Mehta R, Tsimpanakos I. Ten pieces of free software every doctor should have. Lancet. 2007;369:7. Adaptive Biotechnologies. Press Release. Adaptive Biotechnologies announces partnership with Microsoft to decode the human immune system to improve the diagnosis of disease. January 4, 2018. www.adaptivebiotech.com/news/adaptive-biotechnologies-announces-partnership-microsoft-decode-human-immune-system-improve. Accessed January 28, 2018.8. Williams B. Healthcare Blog. Enabling better healthcare with artificial intelligence. August 28, 2017. http://usblogs.pwc.com/emerging-technology/ai-in-healthcare/. Accessed January 28, 2018.9. Winey T. Garbage in, garbage out: avoiding the common pitfalls of AI in healthcare. Becker’s Health IT & CIO Review. June 12, 2017.
To comment on this article, contact rdavidson@uspharmacist.com.
