Blog: Is Artificial Intelligence the future of Health Analytics? *
As the health care industry focuses on the benefits of Artificial Intelligence, medical students and practitioners must not lose sight of how technology is shaping the future of the profession.
If the goal of AI is to create a superhuman intelligence that eclipses that of the most gifted researchers and clinicians, how will the evolution of health analytics condition the means, goals, and ethics of the Life Sciences? Artificial Intelligence is at the core of a major evolutionary transition that is progressively blurring the distinctions between biology, technology, and society. This interdependence challenges the traditional boundaries of mortality, morbidity, and healthcare. After quickly outlining the opportunities of health analytics today, let’s review some of the challenges facing medical practitioners in the immediate future.
Health analytics harnesses data science to develop insights from patterns and correlations in medical data to improve decision-making. Data Science is being leveraged in diagnosing medical conditions, improving the precision of microsurgical procedures, targeting treatment plans, and developing more effective pharmaceuticals. The progressive introduction of electronic health records and virtual assistants is streamlining medical processes and improving both patient safety and outcomes.[i] Current research claims that Machine learning reduces the misdiagnosis of malignant tumors by up to 85%.[ii] The systematic use of algorithms can significantly reduce fraud in medical insurance and payment schemes that currently cost the industry $80 billion annually.[iii] As the focus of the discipline shifts from to prediction to prescription, the economic model of the industry may well be transformed from “cost care” to “health care” in encouraging patients, families, and communities to make lifestyle changes for ensure their future well-being.
Advances in health analytics are modifying our understanding of both mortality and morbidity. Futurists seem more at ease debating whether “transhumans” will live to be several centuries old than exploring what it means to be human.[iv] In our view, one defining characteristic of humanity is autonomy: our capacity to make informed, uncoerced decisions. A second characteristic is agency, the capacity of individuals to act independently. The third is empathy — the capability to understand and to relate to the world around us. The fourth is ethics — shared values that help humanity differentiate right from wrong. Finally, intelligence evokes our ability to acquire and apply knowledge and skills. How will health analytics condition how the medical profession defines well-being?
Health analytics modifies the systemic relationships between patients, physicians, and health institutions. The Quantified Self movement, or “self-knowledge through numbers”, suggests that individuals have the primary responsibility for improving their own physical, mental and emotional well-being. Although the notion of self-tracking has philosophical roots stretching back to at least Pythagoras of Samos, the development of Data Science has largely contributed to the idea that patients can use their personal data to improve “care of oneself”.[v] Recently, the widespread adoption of wearable (the Fitbit, the Apply Watch, UV Sense…) and ambient technologies (Smart Pills, Virtual Voice Assistants, Olfactory technologies…) has fueled this trend.[vi] The application of the resulting data in linear regression models, visualization techniques and data storytelling enables patients to claim a voice in the discussion of their own well-being. Should the medical profession focus solely on the patients they treat or be incentivized for the well-being of the populations they serve?
Although technology’s impact on medical practice has never been neutral, the progressive introduction of artificial intelligence will test the qualifications of the medical profession. The variety and sophistication of algorithms available for descriptive and prescriptive analytics grow exponentially as machine learning applications in the life sciences mature. [vii] Both hospitals and physicians can now leverage AI diagnoses without having to bear the costs and time constraints of retaining more trained professionals. In normalizing the use of machine learning, the practitioner may never take the time to study the model, the code, nor the training data used to establish diagnostic thresholds. Even more disturbingly, the profession’s ever-increasing reliance on data has subtly modified the traditional definitions of “well-being”, “confidentiality”, “truthfulness” and “trust”.[viii] To what extent does the medical profession need to understand how AI changes medical practice?
Algorithms learn by processing past experience: these rule-based procedures implicitly reproduce the human biases that condition human decision-making. Health analytics is currently crossing the threshold from decision support systems that help medical professionals take decisions to the realm of artificial intelligence where algorithms will be designed to replace human decision-making. Machine learning algorithms rely heavily on profiling which uses aggregates of personal data to evaluate predict in specific contexts and situations. These profiles reflect several human biases including functional fixedness, design fixation, and analogy blindness that condition medical procedures of institutions, teams, and practitioners.[ix] If institutions and individuals can be held accountable for their acts, who will be ultimately held responsible for the implicit bias of artificial intelligence?
Current applications of “limited” artificial intelligence won’t fuel innovation in the Life Sciences. Process, produce or service innovation depends upon applying invention effectively in new contexts and in new conditions. AI learns from example; the algorithms are built to test features of variables that can be empirically measured. Even in cases where pathologies are sufficiently specified, AI can’t explore all the possible features (the Obscure Features Hypothesis) that could produce innovative solutions to the challenge at hand.[x] Though AI mimics rational thinking, it proves of little value in replicating other forms of human intelligence: emotional (interpersonal), linguistic (word smart), intrapersonal (self-knowledge) or spiritual (existential) that influence the doctor/patient relationship.[xi] Which types of intelligence will be vital to future innovation in healthcare?
In sum, Artificial Intelligence alone poorly illuminates the future of health analytics. Data Science has already demonstrated in value in improving medical imagery, targeting treatment plans, and accelerating the development of new pharmaceuticals. AI can potentially help practitioners and researchers alike recognize implicit biases that lead to diagnostic errors and ill-adapted medical practices. This said the medical profession needs to understand how Data Science is changing the nature of health care. Future innovation will depend upon mixing and matching human and machine expertise to recognize the contributions of multiple forms of intelligence. The medical profession must look beyond the limits of AI to encourage practitioners to do more, rather than less, in developing our future well-being.
* This contribution will provide the framework of our keynote discussion May 15th for the Deusto conference on “Creating value in healthcare through innovation management “ in San Sebastián.
Lee Schlenker is a Professor of Business Analytics and Community Management, and a Principal in the Business Analytics Institute http://baieurope.com. His LinkedIn profile can be viewed at www.linkedin.com/in/leeschlenker. You can follow the BAI on Twitter at https://twitter.com/DSign4Analytics
[i] Evans, R.S. (2016), Electronic Health Records: Then, Now, and in the Future
[ii] Kontzer, T., (2016), Deep Learning Drops Error Rate for Breast Cancer Diagnoses by 85%
[iii] Sennaar, K., (2019), How America’s Top 4 Insurance Companies are Using Machine Learning
[vi] Lee, J. et al. (2018), Holistic Quantified Self Framework for Augmented Human
[viii] Mittelstadt, B. and Fioridi, L, (2016), The Ethics of Big Data, Current and foreseeable Issues in the biomedical contexts
[ix] Saposnik, G. et al. (2016), Cognitive biases associated with medical decisions: a systematic review
[x] McCaffrey, T. and Spector, L. (2012), Behind every innovative solution lies an obscure feature
[xi] Martin, (2017), Types of Intelligence and How to Find The One You Are Best In