Project Syndicate (USA): “peplomycin” from medical technology

LONDON. In an interview with the Wall Street Journal, published earlier this year, David Feinberg, head of Google Health and self-styled astrologer, enthusiastically said, “If you believe me that we only systematize information in order to make life easier to your doctor, I’m allowing myself some paternalism is hardly ever gonna give that up”.

In other words, patients soon will have no choice but to clinical personalized horoscopes, based on their own medical history and the findings on the basis of growing the Bank’s history. But even if we want to live in such a world, we must look closely at what actually sell current adherents of medical technology.

In recent years, the most American of Big Tech companies — as well as many startups, companies, Big Pharma and others — have entered the field of medical technology. Using big data Analytics, artificial intelligence (AI) and other new methods, they promise to reduce the cost of the troubled health care system, to revolutionize the field of physicians making treatment decisions — in short, to save us from ourselves. What can go wrong?

It turns out, a lot. In the book Weapons of Math Destruction (“Weapons of math destruction”) data analyst Cathy O’neil gives many examples of how algorithms and data without any malicious intent to drive. When the transparent algorithms for the feedback data applied to baseball, they worked even better than expected; but when such models are used in Finance, insurance, law enforcement and education, they can be very discriminatory and destructive.

Health in this sense is no exception. The medical data of the people affected by subjective clinical decisions, medical errors and the evolution of current practice, and the quality of large sets of data often deteriorates because of the lack of records, measurement errors and lack of structure and standardization. However, the big data revolution in health care is presented as if these troubling restrictions exist. Worse, it carried a lot of responsible decision-makers in the field of medicine.

One could argue that if the new solutions provide some advantages, they deserve attention. But without large-scale, well-designed empirical studies we can confidently say whether data Analytics and AI to improve the current situation. And it lacks not only evidence of this benefit — there is no infrastructure or regulatory framework to collect them. Applications for big data is simply introduced in the health care system as if they were harmless or truly useful.

Consider Project Nightingale, private exchange of data between Google and Ascension Health, a large nonprofit health system in the United States. When in November last year, the Wall Street Journal first reported about these secret contacts, it caused a scandal because of concerns about patients ‘ personal data. Worse, as openly admitted Feinberg in the same newspaper, just two months later, “we didn’t know what to do.”

Given that Big Tech companies have no experience in the field of health, such recognition should not be surprising, despite attempts to convince us otherwise. Moreover, while privacy for a person becomes more of a luxury than a right, the algorithms that increasingly govern our lives, turn into inaccessible black boxes, closed from public or regulatory controls to protect corporate interests. And in the case of algorithmic models of healthcare diagnosis and decision making sometimes give results that are not understand the doctors themselves.

Despite the fact that many of those who seek in health, have good intentions, the current approach in the industry based on a lack of ethics and lack of awareness. No one objects to improving health through technology. But before entering into partnerships with technology companies, leaders, and healthcare organizations need to better understand the medical technology.

To begin with, what is important to remember that insights based on big data are made with the help of statistics and mathematics, which requires specific literacy. When the algorithm detects a “causal link” or some other basis of Association, this information may be useful for further research based on hypotheses. But when it comes to actually making decisions, mathematically driven predictive models are not more reliable than the data entered into them. And because their fundamental assumptions are based on what is already known, they give us a picture of the past and the present but not the future. Such applications have great potential to improve health care and reduce costs, but the benefits are not guaranteed.

Another critical area is the AI, which requires as its own architecture — that is, the rules and the underlying logic that determines how the system works, and the access to huge volumes of potentially sensitive data. The goal is to configure the system so that it could itself be trained to find an optimal solution to problems. But here we must remember that the creators of the architecture — people who write the rules and formulating problems is as biased as everyone else, regardless of whether they want to or not. Moreover, as in the case of data analysis, artificial intelligence systems are guided by data from the health care system in its current form, which may lead to playback them, and her failures, and her successes.

Thus, the improvement of health care with big data and AI will probably require much more trial and error than I think theoptimist. Big data projects if they are done transparently and publicly, can teach us to create high quality data sets for the future, thereby increasing the chances of success of the algorithmic solutions. In addition, must be available and the algorithms — if not publicly, then at least for the regulators and associations who subscribe to the service.

First and foremost, healthcare organizations and governments need to open your eyes and critically think about the consequences of new health care programs, largely unchecked. Instead of just handing out medical history of patients and other data, hospitals and regulators should monitor the developers of the technology sector, which builds architecture and implement new experimental system. More people should Express their opinions and critically assess the assumptions underlying the initial prototypes, and after this you need to conduct controlled experiments to evaluate real-world performance of these technologies.

Healthcare solutions based on big data being over-hyped, hastily sent to market without much regulation, transparency, standardization, accountability or reliable methods of validation. Patients deserve to the health system and medical staff have protected them, and not just used as data sources for experiments aimed at obtaining profit.