February 22, 2019

The bioethics of machine clinical decision making

Artificial intelligence (AI) in healthcare and research
Regulation of predictive analytics in medicine

This is what a brief note from Nuffield Council of Bioethics says about artificial intelligence in healthcare:
The use of AI raises ethical issues, including:
  • the potential for AI to make erroneous decisions; 
  • the question of who is responsible when AI is used to support decision-making; 
  • difficulties in validating the outputs of AI systems; inherent biases in the data used to train AI systems; 
  • ensuring the protection of potentially sensitive data; 
  • securing public trust in the development and use of AI; 
  • effects on people’s sense of dignity and social isolation in care situations; 
  • effects on the roles and skill-requirements of healthcare professionals; 
  • and the potential for AI to be used for malicious purposes.
A key challenge will be ensuring that AI is developed and used in a way that is transparent and compatible with the public interest, whilst stimulating and driving innovation in the sector.
This statement is naive.(From m-w, naive:  marked by unaffected simplicity : INGENUOUS). Up to now, have you seen any transparent algorithm available for imaging, triage or any app? For sure not. Therefore, the real key challenge is to stop introducing such algorithms (ban apps) unless there is a regulatory body that takes into account the quality assurance-effectiveness side (sensibility, especificity) and the required transparency for citizens.
Nuffield has released a brief up to now. Let's wait for the report.
If you want a quick answer, check Science this week:
To unlock the potential of advanced analytics while protecting patient safety, regulatory and professional bodies should ensure that advanced algorithms meet accepted standards of clinical benefit, just as they do for clinical therapeutics and predictive biomarkers. External validation and prospective testing of advanced algorithms are clearly needed
 They explain the five standards and give rules and criteria for regulation. It is really welcome.



February 21, 2019

Pharm niche busters

The Information Pharms Race and Competitive Dynamics of Precision Medicine: Insights from Game Theory
Economic Dimensions of Personalized and Precision Medicine
Precision medicines inherently fragment treatment populations, generating small-population markets, creating high-priced “niche busters” rather than broadly prescribed “blockbusters”. It is plausible to expect that small markets will attract limited entry in which a small number of interdependent differentiated product oligopolists will compete, each possessing market power.
A chapter in a new book on  Precision Medicine explains the new approaches to a oligopolistic market structure where the size of the market may be determined by biomarkers with cut-off valued suggested by pharmaceutical firms. The dynamics of this market is described according to game theory.
I already have pending chapters to read of this book. A must read.



February 16, 2019

Defining roles and skills for digital health

The Topol Review
Preparing the healthcare workforce to deliver the digital future.

The NHS asked Dr. Eric Topol about the new health workforce and how digital health will change the current landscape. A must read:
This is an exciting time for the NHS to benefit and apitalise on technological advances. However, we must learn from previous change projects. Successful mplementation will require investment in people as well s technology. To engage and support the healthcare workforce in a rapidly changing and highly technological orkplace, NHS organisations will need to develop a learning environment in which the workforce is given very encouragement to learn continuously. We must better understand the enablers of change and create culture of innovation, prioritising people, developing an agile and empowered workforce, as well as digitally capable leadership, and effective governance processes
to facilitate the introduction of the new technologies, supported by long-term investment.

February 15, 2019

Who is worse off?

Health, priority to the worse off, and time

The prioritisation of resource allocation towards the worse off is a well known rule. What does this mean exactly?
 There are many dimensions in which someone can be worse off (e.g., in terms of wellbeing, health, opportunities, resources), and there are many ways to give priority to someone (e.g., by giving extra weight to their claims, lexical priority to their claims, or by earmarking a fixed amount of resources for their claims). Furthermore, there are many different reasons why one might want to give priority to benefits to the worse off: is it because it is good to promote equality for its own sake, good to promote equality for other reasons, because benefits to the worse off matter more, because the worse off typically fall under some sufficiency threshold, or for many of these (and maybe other) reasons
The precise argument is described in a recent article that combines the complete lives approach with the forward looking approach, and says:
 I believe that the focus on complete lives has been beneficial in that it is a step away from a complete focus on current distributions of health. However, I think that the arguments presented in this paper give us reason to adopt a more nuanced approach to how to rank individuals in terms of who is worse off with the purpose of giving priority to certain benefits in light of unequal distributions of health over time. Such an approach accepts that both the complete lives view and the forward looking view that only takes into account current and future health states, matter. This leads to the complicated question of how to combine these views. Some work that addresses how to combine  concerns for simultaneous segment inequality and complete lives inequality has appeared recently, but the question needs further attention.
Therefore, it is still a work in progress.


February 8, 2019

The perfect storm of surveillance capitalism


THE DEFINITION
1. A new economic order that claims human experience as free raw material for hidden commercial practices of extraction, prediction, and sales;
2. A parasitic economic logic in which the production of goods and services is subordinated to a new global architecture of behavioral modification;
3. A rogue mutation of capitalism marked by concentrations of wealth, knowledge, and power unprecedented in human history;
4. The foundational framework of a surveillance economy;
5. As significant a threat to human nature in the twenty-first century as industrial capitalism was to the natural world in the nineteenth and twentieth;
6. The origin of a new instrumentarian power that asserts dominance over society and presents startling challenges to market democracy;
7. A movement that aims to impose a new collective order based on total certainty;
8. An expropriation of critical human rights that is best understood as a coup from above: an overthrow of the people’s sovereignty.
Last month in my post on the book: Modern monopolies I wanted to highlight the current trend towards monopolies using platforms as a business model. Now you may add a complementary perspective with the book: The age of surveillance capitalism. While the former emphasizes the business perspective, the later focus on behavioral prediction surplus and how it is generated. It provides a social perspective of the current "surveillance capitalism". In my opinion there is a lot of current economy that already confirms this view, it is not a future expectation.
Our lives are rendered as behavioral data in the first place; ignorance is a condition of this ubiquitous rendition; decision rights vanish before one even knows that there is a decision to make; there is no exit, no voice, and no loyalty, only helplessness, resignation and psychic numbing; encryption is the only positive action left to discuss.
Surveillance capitalists take command of the essential questions that define knowledge, authority, and power in our time: Who knows? Who decides? Who decides who decides? 
As you may imagine, this is a book that once you started it's impossible to stop reading. Highly recommended if you want to understand current hot topics and social trends.
A perfect storm is an event in which a rare combination of circumstances drastically aggravates the event. This is exactly what we have right now in front of us, and as we are inside the wave we are not able to recognise what's going on.

February 7, 2019

Medicine as a data science (5)

A guide to deep learning in healthcare

Some months ago, Mckinsey released a guide to AI for executives. It says:
Deep learning is a type of machine learning that can process a wider range of data resources, requires less data preprocessing by humans, and can often produce more accurate results than traditional machine-learning approaches (although it requires a larger amount of data to do so). In deep learning, interconnected layers of software-based calculators known as “neurons” form a neural network. The network can ingest vast amounts of input data and process them through multiple layers that learn increasingly complex features of the data at each layer. The network can then make a determination about the data, learn if its determination is correct, and use what it has learned to make determinations about new data. For example, once it learns what an object looks like, it can recognize the object in a new image.
Now Nature publishes a helpful review article on deep learning in healthcare.
Some of the greatest successes of deep learning have been in the field of computer vision (CV). CV focuses on image and video understanding, and deals with tasks such as object classification, detection, and segmentation—which are useful in determining whether a patient’s radiograph contains malignant tumors
The next step is speech and text. Some advances are already available. Basically, Tensorflow by Google is feeding the beast.

PS. WHO and the classification of digital health interventions 1.0

PS. And the book to read:


February 4, 2019

When the regulator doesn't care about the danger within us

A must see Netflix documentary: The bleeding edge. It explains how medical devices are introduced in the market without appropriate control.
CBS news explains some details:


Just because it's new doesn't mean it's better, it may be dangerous and damage you for life. Unfortunately, this is the summary.
And the book to read:


February 2, 2019

Medicine as a data science (4)

The practical implementation of artificial intelligence technologies in medicine

One of the critical issues for AI implementation in clinical practice is about privacy. In this article you'll find a clear statement on the impact of EU regulation:
The GDPR will affect AI implementation in healthcare in several ways. First, it requires explicit and informed consent before any collection of personal data. Informed consent has been a long-standing component of medical practice (unlike in social media or onlinebased marketing), but having to obtain informed consent for an  collection of data still represents a higher bar than obtaining consent for specific items, such as procedures or surgical interventions. Second, the new regulation essentially lends power to the person providing the data to track what data is being collected and to be able to request removal of their data. In the healthcare context, this will shift some of the power balance toward the patient and highlights the importance of ongoing work needed to protect patient privacy and to determine appropriate governance regarding data ownership. 
More details inside.

Potential roles of AI-based technologies in healthcare.


 Integration of patient health information at multiple interfaces.







February 1, 2019

Medicine as a data science (3)

High-performance medicine: the convergence of human and artificial intelligence

If you want to know the current state of artificial intelligence in medicine, then Eric Topol review in Nature is the article you have to read. A highlighted statement:
There are differences between the prediction metric for a cohort and an individual prediction metric. If a model’s AUC is 0.95, which most would qualify as very accurate,
this reflects how good the model is for predicting an outcome, such as death, for the overall cohort. But most models are essentially classifiers and are not capable of precise prediction at the individual level, so there is still an important dimension of uncertainty.
And this is good summary:
Despite all the promises of AI technology, there are formidable obstacles and pitfalls. The state of AI hype has far exceeded the state of AI science, especially when it pertains to validation and readiness for implementation in patient care. A recent example is IBM Watson Health’s cancer AI algorithm (known as Watson for Oncology). Used by hundreds of hospitals around the world for recommending treatments for patients with cancer, the algorithm was based on a small number of synthetic, nonreal cases with very limited input (real data) of oncologists. Many of the actual output recommendations for treatment were shown to be erroneous, such as suggesting the use of bevacizumab in a patient with severe bleeding, which represents an explicit contraindication and ‘black box’ warning for the drug. This example also highlights the potential for major harm to patients, and thus for medical malpractice, by a flawed algorithm. Instead of a single doctor’s mistake hurting a patient, the potential for a machine algorithm inducing iatrogenic risk is vast. This is all the more reason that systematic debugging, audit, extensive simulation, and validation, along with prospective scrutiny, are required when an AI algorithm is unleashed in clinical practice. It also underscores the need to require more evidence and robust validation to exceed the recent downgrading of FDA regulatory requirements for medical algorithm approval

Therefore, take care when you look at tables like this one:



PredictionnAUCPublication (Reference number)
In-hospital mortality, unplanned readmission, prolonged LOS, final discharge diagnosis216,2210.93* 0.75+0.85#Rajkomar et al.96
All-cause 3–12 month mortality221,2840.93^Avati et al.91
Readmission1,0680.78Shameer et al.106
Sepsis230,9360.67Horng et al.102
Septic shock16,2340.83Henry et al.103
Severe sepsis203,0000.85@Culliton et al.104
Clostridium difficile infection256,7320.82++Oh et al.93
Developing diseases704,587rangeMiotto et al.97
Diagnosis18,5900.96Yang et al.90
Dementia76,3670.91Cleret de Langavant et al.92
Alzheimer’s Disease ( + amyloid imaging)2730.91Mathotaarachchi et al.98
Mortality after cancer chemotherapy26,9460.94Elfiky et al.95
Disease onset for 133 conditions298,000rangeRazavian et al.105
Suicide5,5430.84Walsh et al.86
Delirium18,2230.68Wong et al.100