The Decision Link AI-Cloud Framework for Personalized Medicine
A FRAMEWORK FOR USING AI-CLOUD TO TURN INFORMATION INTO BETTER ACTIONS AT SCALE FOR CONTINUALLY-IMPROVING PERSONALIZED MEDICINE OUTCOMES
December 12th, 2023
Despite over $5 billion invested in AI for medical imaging and an estimated $150 billion dollars in cost savings in the US alone by 2026, our healthcare systems continue to spiral downwards into a doom loop of failure.¹ ²
The US has the worst outcomes of any high-income nation at a staggering $4T pricetag.² It’s a system with massive inequities across patient populations.³ Radiology is a key hub, making key decisions for ~80% of patients and ~90% of health data.⁴ ⁵ Today, radiologists are quitting in droves, further accelerating the doom loop.⁶
We believe the root cause for this continued failure is the breaking “decision link” which causes declining team and individual decision performance. This core problem is not properly recognized, tracked, measured, or analyzed. Here, we define the “decision link” and explain why it’s breaking. We provide a novel Decision Link AI-Cloud Framework to help eliminate failure points and help clinicians, especially radiologists, to guide successful AI-cloud adoption.
Through stewardship of novel Decision Augmentation and Decision Intelligence innovation with AI-cloud, this framework will enable physicians to drive radically-improved decision quality and efficiency in fast-paced and high-stakes clinical environments, with far less stress and cognitive overload. The secondary potential positive impacts include improved system-level learning, lower costs, and continually-improved personalized patient care outcomes.
The Critical Human Decision Link (Figure 1)
Decision-making connects data, analytics, and predictions to best actions to generate a desired outcome.⁷ ⁸ ⁹ ¹⁰ For example, the military uses the OODA Loop (Observe, Orient, Decide, Act) framework where a final decision-point leads to an action step. According to the discipline of “decision engineering,” a decision-making process leads to an action with an “irrevocable allocation of resources.” ⁸ For healthcare, it’s appropriate to be more specific and define it as an irrevocable allocation of resources that balances risks versus benefits for a patient.
We define the Decision Link as the activity of making decisions that answer, “What should I do?” leading to actions with an irrevocable use of resources, and all systems-level factors that impact these decisions, including human, technical, data, and analytics factors.
For this critical link in the chain of patient care, we define decision performance as having three primary elements: decision quality, decision efficiency, and the capacity for learning.¹² Healthcare’s expert decision-makers must be able to make high quality, high-stakes decisions quickly and efficiently. Professionals need systems which enable learning from patient care decisions to improve their decision-making over time.
The Elephant in the Room — AI Can’t Safely Replace Human Decision-Makers
Before going any further, let’s take on the elephant in the room. With all the hype in recent years suggesting AI can replace human physician decision-makers, like radiologists, there are many who will believe key decision-makers can be computers instead of trained human experts.
Human medical decision-making is far more robust and “transferable” than brittle AI algorithms. Humans easily handle unique cases, gaps in data, and situations demanding nuanced contextual understanding.¹² ¹⁸ ¹⁹ Neural networks powering “AI” are powerful classifiers, providing useful predictive information, yet AI algorithms are very brittle.¹² ¹³ ¹⁴ ¹⁵ ¹⁶ It’s not surprising clinical studies show AI brittle failure rates of up to 93% despite “super human” classifications “in the lab.” ¹⁷ While AI in healthcare has the capacity to make very simple decisions automatically, it does not perform well in complex and chaotic decision spaces, where humans excel.¹¹ While most AI algorithms cannot be easily transferred from one practice location to another due its brittleness, the trained human — such as a radiologist — can perform at a high level in any location on a myriad of highly variant datasets.
Beyond this technical viewpoint, we also argue on ethical grounds that patient guidance and oversight are an essential fixture in healthcare. Patients must have professionals to advocate for their best interests — as it has been for thousands of years since Hippocrates. Physicians, nurses, and other healthcare professionals have the capacity to drive decisions for others’ benefit based on emotional intelligence, human judgement, and, simply, the capacity for caring.
We conclude that the human “decision link” must be the decider of what actions are taken in virtually all patient care cases. Our new powerful AI tools should be used in healthcare systems to augment human decision-making capabilities for “graceful extensibility.” ¹⁹ ²¹ ²² ²³
Let’s now deepen this argument through the lens of medical history and how many medical decisions are made today — namely, with prescriptive analytics.
Medicine — The Inventors of Prescriptive Analytics
Medicine invented “prescriptive analytics,” which answers the question, “What should I do?” It suggests options for mitigating risks and reaching a desired outcome, and provides metrics for each option.²⁴ ²⁵ Physicians widely use “decision engineering” and clinical science for prescriptive analytics today at innumerable points-of-care for patients across all medical specialties.
For example, BI-RADS guidelines help radiologists choose whether to biopsy a lesion seen on breast cancer screening imaging versus surveillance imaging, etc.²⁶ The radiologist decides based on imaging features and clinical data, yet also on other patient specific contextual considerations. A biopsy may not be pursued, for example, if there is an existing co-morbidity that increase potential complications. Guidelines also exist in every other field of medicine. The human decision-maker adjusts based on a wide array of considerations for each patient.
How does medicine design each prescriptive analytics use case? It uses “decision engineering” (a term now used in the business realm) to map decisions to actions and outcomes using clinical research. ²⁴ ²⁷ ²⁸ Over hundreds of years, medicine has developed its clinical sciences so that physicians can answer the question, “what should I do?” at the point-of-care for each patient.
The Breaking Decision Link Problem (Figure 2)
Despite this long history of excellence in clinical medicine around high quality decision-making, toxic technology designs today are breaking this chain.
Today, poor human-information interaction designs and an imbalanced fixation on gathering excess data, then embellishing that data further using AI, is breaking the vital “decision link.”
“The amount of data that healthcare churns out with each passing second is almost incomprehensible.” ²⁹
Healthcare decision-makers are “drowning in data and starving for information.” ³⁰ Increasing complexity related to increasing volumes, veracity, velocity, and variety (the “4 V’s”) of big data is exploding. The compound annual growth rate for health data is 36%, the fastest of any domain. Medical imaging volumes have increased 10X in the last 20 years, and the increase in decision complexity across all domains is on average estimated at a 65% increase in the last 2 years. ³¹ ³² ³³ ³⁴ ³⁵ ³⁶ A fixation among investors on AI “point solutions” using massive data is worsening the information overload and complexity for human experts at points-of-care.³⁷ ³⁸ ³⁹ Further, unlike virtually every domain, ~80% of health data is unstructured, meaning there’s limited ability to filter and query the data to create efficiencies.⁴⁰ Had we also assured high quality user experience designs for physicians managing the four V’s of all this big data, then we would likely see far more success than the latent AI adoption we see today.⁴¹
Healthcare technology vendors have given little consideration to physician user experience, instead focused by IT engineers on machine processes and add-on “workflows.” The result is interface chaos with fragmented data sources, where physicians search for, retrieve, and view each and every piece of individual raw data on legacy viewers designed in the 1980’s. This slowed, painstaking process creates enormous stress in clinical environments. Based on a small survey (corroborated by other vast interview data) of radiologists, many unmeasured factors impact radiologist decision performance. For example, radiologists responded that stress (100%), high stakes (83%), decision complexity and data overload (50%), missing or hard to find information (90%), RVU (a metric for report output productivity) and time pressures (73%), distractions (100%), and high cognitive effort (82%) negatively impact their decision-making.⁴² Radiologist burnout (54%) and moral distress (98%) are at an all-time high.⁴³ ⁴⁴ These physicians hold a critical role in the “decision link,” making key decisions for ~80% of patient cases, yet are quitting in droves.⁵ ⁶
We are in a doom loop where poor physician user experience in fast-paced clinical environments leads to slower decisions, escalating data chaos, and error-prone poor decisions which cyclically worsens physician experience.
The result is an alarming decline in individual and team decision performance. The National Academies estimated that diagnostic error rates in the US are as high as 10%-20%.⁴⁵ This shocking data is corroborated by more recent data from other groups, such as the nation’s largest employer, Walmart, using data from their employer-sponsored insurance programs.⁴⁶ Based on survey data, radiology quality and safety professionals say gathering decision performance data (100%) and deciphering which AI solutions improve patient care outcomes (86%) are painful.⁴⁴ It’s clear that this doom loop has led to worsening decisions, worsening patient care outcomes, and escalating operational costs that are not properly tracked, measured, or analyzed.
The results from the ignored “decision link” are truly tragic. While physicians may be suffering, it’s the patients and society who bear the ultimate burden. The US has the world’s worst outcomes of any high-income nation with healthcare costs bankrupting our middle class and schools. In fact, medical bills are the #1 cause of bankruptcy in the US.⁴⁷ ⁴⁸
We face a massive and worsening crisis.
It’s Time to Augment the Decision Link (Figure 3)
It’s essential that we recognize the vital importance of the “decision link” in healthcare and how we can augment the chain with new technology.
So, here’s the good news.
We can fix the breaking “decision link” by using proven paradigms that exist in other domains. In fact, we can use new technology to augment the decision link. We can take medicine’s prescriptive analytics into our AI-cloud-enabled future by leveraging new decision sciences, as well as new tools, methods, and resources. We have created a framework to put these key puzzle pieces together to empower healthcare professionals to guide successful clinical adoption of AI-cloud technology.
The Decision Link Framework is a guide for using AI-cloud, responsibly, to turn information into better actions at scale for continually-improving personalized medicine outcomes.
This framework has four layers and eleven components centered around augmenting and optimizing decisions at the “decision link.” Components include:
Human-Centered Design: Apple products exemplify human-centered design processes by having these key elements: designs are centered on humans first, designs solve fundamental basic problems, and designing requires systems-level thinking to produce a global (not local) optimization.⁴⁹ This is a critical concept for IT purchasers to understand in healthcare today. In radiology, for example, virtually all new vendors latch AI algorithms onto existing known PACS user “workflows.” Some have created PACS with AI-cloud components using modernized IT architectures, yet did so for what is essentially the very same user interface viewer designs unchanged since the 1980’s. They offer no new core user capabilities.⁵⁰ Very few companies have pursued a truly human-centered design process to re-invent the physician (radiologist) experience.⁵⁰ This concept is also important for investment decisions in AI-cloud, as successful human-centered designs solve major problems and thus provide a leap forward in user and customer value. It differentiates investing in a “red ocean,” such as the entrenched legacy PACS $5 billion dollar global market, versus the estimated “blue ocean” $88 billion dollar global market for digital health platforms that leverage medical imaging.
Decision Augmentation. AI-cloud enables novel experience designs to augment users’ decision-making by properly considering human factors. In Decision Augmentation (DA), rich interaction occurs between humans and computers, the human asks the questions and makes complex decisions, and computers may make algorithmic decisions (such as classifying images and generating recommendations on potential actions), yet the human makes the final decision.¹¹ ⁴⁹ Healthcare users can be augmented to increase their decision efficiency by removing many of the burdens of today’s toxic designs, such as limiting time spent gathering data, removing the need to review unimportant data, removing repeated tasks, and by automatically surfacing the key data for each decision.⁵¹ It can also augment decision-makers through Multimodal AI methods which can richly integrate various types of information in the cloud to present more powerful predictions to optimize decision quality.⁵⁰ ⁵¹ (see patents section in references)
Decision Intelligence. Decision Intelligence (DI) is a framework for using AI to turn information into better actions at scale.⁸ This modern definition was coined by Google’s first Chief Decision Scientist; the original ideation and term can be traced to MIT articles from the late 1960’s.⁸ ⁹ ¹⁰ DI takes the “best routes” prescriptive analytics that physicians use today into our AI-powered future. A widely used example of DI in action is Google Map’s “best routes.” Other uses include Formula 1 racing and mortgage approvals. DI is enabled by new AI-cloud systems that allow more powerful data analytics that can connect all types of data and map to outcomes in real-time, enabling continuous monitoring and improvement.⁵¹ ⁵²
What can we do with DI in healthcare? DI enables increasing personalization of each patient’s care pathway and continuous outcomes improvement. In today’s prescriptive analytics use cases, physicians at the point-of-care today may place a given patient in one of, say, four categories based on after-the-fact clinical research using available (yet hard to access and organize) population data. With more modern AI-cloud systems, the data and analytics already exist in the cloud and can be updated in real-time. This new technology represents a major opportunity. With DI, physicians will be able to see far more potential “best routes,” perhaps thousands, at the point-of-care which are far more personalized in order to drive better optimized outcomes for each patient. DI is currently at the earliest stages in healthcare, with only one known clinical DI vendor and a handful of operations DI vendors.⁵¹ ⁵³
Personalized Medicine Prescriptive Analytics. DI will take what medicine already does extremely well, namely prescriptive analytics, into the AI-cloud future by allowing us to consider far more complex information about each patient. DI systems of tomorrow will help decision-makers better leverage information that we know is vital, but not yet included in guidelines. For example, it has been well-established that social determinants of health such as living, education, and personal finances often have a larger impact on patient care outcomes than other factors.⁵⁴ Physicians and other clinical researchers can forge more powerful prescriptive analytics use cases that are better designed to help eliminate the vast inequities we see today.
Clinical research and practice case studies. Forward progress requires medicine to do what it has done well for centuries, while also embracing new sciences to properly measure, analyze, and study the impact of technology on the “decision link.” A new focus on decision science in healthcare is urgently needed. “While most fields of research focus on producing new knowledge, decision science is uniquely concerned with making optimal choices based on available information.” ⁵⁵ Visual analytics is a new field in the decision sciences; it’s the science of analytical reasoning facilitated by interactive visual interfaces. This new science was established as a response to similar cognitive overload problems in the intel community, blamed for the US missing key intelligence leading up to 911.⁵⁶
Our Human(e) Future
Who will drive positive change for a responsible AI-cloud for our future healthcare system? The same humanitarians who successfully drove responsible science and progress over the last thousands of years, namely practicing healthcare professionals. Our future will be built by well-intentioned human experts who both deeply understand the clinical context and care enough to power the future without allowing AI to take the wheel.
Clinical experts and scientists, including radiologists, have an essential expertise for augmenting the “decision link.” They do not need to be deep technical or data science experts to guide optimized decision-making. By applying the Decision Link Framework to innumerable potential medical use cases, professionals will be empowered to address the crisis-level problems of today and drive our future of personalized care.
Why should you care? Our healthcare crisis impacts every family, including yours.
It’s time for the productive phase of the hype cycle. It’s time for smart digital health investments. It’s time to meet crisis with success.
This is our future. This is our value-based care future.⁵⁷
This is our human-enabled future.
The author extends a special thank you to Roy Shulte, a recognized foremost leader in Decision Intelligence and former Distinguished VP Analyst at Gartner (retired), for his expert guidance during development of this Decision Link Framework.
About the Author: Dr. Moira Schieke is the Chief Medical Officer/CEO and Founder of Cubismi, a digital medicine innovator and pioneer, cancer imaging and MRI expert, and a board-certified clinical radiologist. She is the chief designer of the Decision Flow-3D™ AI-cloud platform, as well as chief inventor of its patent-protected Multimodal AI / digital twin technology. Dr. Schieke is an advocate for patient digital rights and for breaking down legacy barriers to improved care. She is also an accomplished professional artist and figure painter.
References:
Select Cubsimi Published Patents:
1. Artificial Intelligence in Healthcare | Accenture. https://www.accenture.com/au-en/insights/health/artificial-intelligence-healthcare.
2. Investment in Medical Imaging AI Tops $5B. Signify Research https://www.signifyresearch.net/medical-imaging/investment-in-medical-imaging-ai-tops-5b/.
3. U.S. Health Care from a Global Perspective, 2022 | Commonwealth Fund. https://www.commonwealthfund.org/publications/issue-briefs/2023/jan/us-health-care-global-perspective-2022.
4. The U.S. spends nearly $4 trillion on health care, but inequities still exist. Here’s why. PBS NewsHourhttps://www.pbs.org/newshour/show/the-u-s-spends-nearly-4-trillion-on-health-care-but-inequities-still-exist-heres-why (2021).
5. HealthManagement.org, margarettacolangelo, RealMargaretta, dmitrykaminskiy & DeepTech2_0. Radiology Management, ICU Management, Healthcare IT, Cardiology Management, Executive Management. HealthManagementhttps://healthmanagement.org/c/hospital/issuearticle/ai-in-medical-imaging-may-make-the-biggest-impact-in-healthcare (2023).
6. Smith-Bindman, R., Miglioretti, D. L. & Larson, E. B. Rising Use Of Diagnostic Medical Imaging In A Large Integrated Health System. Health Aff. Proj. Hope 27, 1491–1502 (2008).
7. Times Are Tight: Staff Shortages Prompt New Strategies • APPLIED RADIOLOGY. https://appliedradiology.com/articles/times-are-tight-staff-shortages-prompt-new-strategies.
8. Link | How Decision Intelligence Connects Data, Actions, and Outcomes for a Better World — Link. https://www.lorienpratt.com/linkthebook/.
9. The steering wheel for your life — Decision Intelligence Video Tutorial | LinkedIn Learning, formerly Lynda.com. LinkedIn https://www.linkedin.com/learning/decision-intelligence/the-steering-wheel-for-your-life.
10. Poensgen, O. H. The information system: data or intelligence? MIT Press (1969).
11. White, L. S. (Leon S. Systems analysis and management decision-making. (1971).
12. Decision Performance — The metric that predicts if you will outperform or just be average. (2021) https://www.linkedin.com/pulse/decision-performance-metric-predicts-you-outperform-dellermann.
13. 7 Revealing Ways AIs Fail — IEEE Spectrum. https://spectrum.ieee.org/ai-failures.
14. Cremer, D. D. & Kasparov, G. AI Should Augment Human Intelligence, Not Replace It. Harvard Business Review(2021).
15. Reader, T. M. P. “Hallucinating” AIs Sound Creative, but Let’s Not Celebrate Being Wrong. The MIT Press Readerhttps://thereader.mitpress.mit.edu/hallucinating-ais-sound-creative-but-lets-not-celebrate-being-wrong/ (2023).
16. Arbel, J. A Primer on Bayesian Neural Networks: Review and Debates. (2023).
17. Wu, E. et al. How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals. Nat. Med. 27, 582–584 (2021).
18. Rebooting AI. Rebooting AI http://rebooting.ai/.
19. Woods, D. First & Last Interview: Boeing 737 Max accidents reveal past results on Automation Surprises. 2019, (2019).
20. Fleishman, H. Comment on the Proceedings of the U.S. Food and Drug Administra- tion (FDA) public workshop regarding the ‘Evolving Role of Artificial Intelligence in Radiological Imaging’ by the American College of Radiology and the Radiological Society of North America. (2020). (2020).
21. Using AI in Decision Making: When and Why. Gartner https://www.gartner.com/smarterwithgartner/would-you-let-artificial-intelligence-make-your-pay-decisions.
22. Nagar, Y. & Malone, T. W. Combining Human and Machine Intelligence for Making Predictions.
23. Woods, D. The Theory of Graceful Extensibility: Basic rules that govern adaptive systems. Environ. Syst. Decis. 38, (2018).
24. Augmented intelligence in medicine. American Medical Association https://www.ama-assn.org/practice-management/digital/augmented-intelligence-medicine (2023).
25. Fang, X., Gao, Y. & Jen-Hwa Hu, P. A Prescriptive Analytics Method for Cost Reduction in Clinical Decision Making. MIS Q. 45, 83–115 (2021).
26. Winters-Miner, L. A. et al. Chapter 1 — History of Predictive Analytics in Medicine and Health Care. in Practical Predictive Analytics and Decisioning Systems for Medicine (eds. Winters-Miner, L. A. et al.) 5–22 (Academic Press, 2015). doi:10.1016/B978–0–12–411643–6.00001–6.
27. Breast Imaging Reporting & Data System. https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Bi-Rads.
28. Poucke, S. V., Thomeer, M., Heath, J. & Vukicevic, M. Are Randomized Controlled Trials the (G)old Standard? From Clinical Intelligence to Prescriptive Analytics. J. Med. Internet Res. 18, e5549 (2016).
29. Noble, B. &. Practical Data Analytics for Innovation in Medicine: Building Real Predictive and Prescriptive Models in Personalized Healthcare and Medical Research Using AI, ML, and Related Technologies|Hardcover. Barnes & Noble https://www.barnesandnoble.com/w/practical-data-analytics-for-innovation-in-medicine-gary-d-miner/1141012453.
30. Culbertson, N. Council Post: The Skyrocketing Volume Of Healthcare Data Makes Privacy Imperative. Forbeshttps://www.forbes.com/sites/forbestechcouncil/2021/08/06/the-skyrocketing-volume-of-healthcare-data-makes-privacy-imperative/.
31. D.Sc, E. D. B. Drowning in Data, Starved for Information. Eric D. Brown, D.Sc. https://ericbrown.com/drowning-in-data-starved-for-information.htm (2014).
32. Skyrocketing Volume of Healthcare Data. Forbes https://www.forbes.com/sites/forbestechcoun- cil/2021/08/06/the-skyrocketing-volume-of-healthcare- data-makes-privacy-imperative/?sh=213007d76555.
33. Healthcare data volume globally 2020 forecast. Statista https://www.statista.com/statistics/1037970/global-healthcare-data-volume/.
34. Top Trends in Data and Analytics for 2021: Engineering Decision Intelligence. Gartnerhttps://www.gartner.com/en/documents/3996986.
35. Socio-economic impact of AI on European health systems. Deloitte Belgiumhttps://www2.deloitte.com/be/en/pages/life-sciences-and-healthcare/articles/the-socio-economic-impact-of-AI-on-healthcare.html.
36. Rosenkrantz, A. B., Hughes, D. R. & Duszak, R. The U.S. Radiologist Workforce: An Analysis of Temporal and Geographic Variation by Using Large National Datasets. Radiology 279, 175–184 (2016).
37. Hosny, A., Parmar, C., Quackenbush, J., Schwartz, L. H. & Aerts, H. J. W. L. Artificial intelligence in radiology. Nat. Rev. Cancer 18, 500–510 (2018).
38. How will AI affect radiologist productivity? AuntMinnie https://www.auntminnie.com/clinical-news/ct/article/15629746/how-will-ai-affect-radiologist-productivity (2021).
39. The radiologist’s gerbil wheel: interpreting images every 3–4 seconds eight hours a day at Mayo Clinic • APPLIED RADIOLOGY. https://appliedradiology.com/articles/the-radiologist-s-gerbil-wheel-interpreting-images-every-3-4-seconds-eight-hours-a-day-at-mayo-clinic.
40. Kong, H.-J. Managing Unstructured Big Data in Healthcare System. Healthc. Inform. Res. 25, 1–2 (2019).
41. AI in Healthcare Is Here, But Uptake Is Slow | HBHI. https://hbhi.jhu.edu/news/ai-healthcare-here-uptake-slow.
42. Radiologist Decision Performance Survey, Conducted by Cubismi. (2023).
43. Radiologists once again rank among the most burned-out specialists. https://healthimaging.com/topics/healthcare-management/medical-practice-management/radiologists-rank-among-most-burned-out.
44. Survey finds moral distress is prevalent among radiologists. AuntMinnie https://www.auntminnie.com/practice-management/article/15633108/survey-finds-moral-distress-is-prevalent-among-radiologists (2023).
45. Improving Diagnosis in Health Care. (National Academies Press, 2015). doi:10.17226/21794.
46. General Session: Lisa Woods and Marcus Osborne, Walmart. (2018).
47. States confront medical debt that’s bankrupting millions. AP News https://apnews.com/article/medical-debt-legislation-2a4f2fab7e2c58a68ac4541b8309c7aa (2023).
48. CEO’s Guide. Health Rosetta https://healthrosetta.org/ceoguide/.
49. What is Human-Centered Design? — updated 2023. The Interaction Design Foundation https://www.interaction-design.org/literature/topics/human-centered-design.
50. Sirona Medical. Sirona Medical https://sironamedical.com/solutions/.
51. Cubismi. https://www.cubismi.com/.
52. Gartner. Know when to augment decisions with artificial intelligence. (2022).
53. Acosta, J. N., Falcone, G. J., Rajpurkar, P. & Topol, E. J. Multimodal biomedical AI. Nat. Med. 28, 1773–1784 (2022).
54. Decision Intelligence Is the Near Future of Decision Making. Gartnerhttps://www.gartner.com/en/documents/4004300.
55. Effective Decision Making Must Be Connected, Contextual and Continuous. https://www.gartner.com/smarterwithgartner/how-to-make-better-business-decisions.
56. Data-driven decision making will fail — and here is why | Computer Weekly. ComputerWeekly.comhttps://www.computerweekly.com/opinion/Data-driven-decision-making-will-fail-and-here-is-why.
57. Social Determinants of Health — Healthy People 2030 | health.gov. https://health.gov/healthypeople/priority-areas/social-determinants-health.
58. Boston, 677 Huntington Avenue & Ma 02115. What is Decision Science? Center for Health Decision Sciencehttps://chds.hsph.harvard.edu/approaches/what-is-decision-science/ (2017).
59. Fisher, B. Illuminating the Path: An R&D Agenda for Visual Analytics. in 69–104 (2005).
60. Blau, S. Utilizing Data and Machine Learning to Change Predictive Analytics into Prescriptive Analytics. (2020).
