News: Diagnostic errors — more harmful, than treatment mistakes.

Recent studies of medical errors have estimated errors may account for as many as 251,000 deaths annually in the United States (U.S)., making medical errors the third leading cause of death. Error rates are significantly higher in the U.S. than in other developed countries such as Canada, Australia, New Zealand, Germany and the United Kingdom (U.K). At the same time less than 10 percent of medical errors are reported.

Source: The BMJ

After reviewing 25 years of U.S. malpractice claim payouts, Johns Hopkins researchers found that diagnostic errors — not surgical mistakes or medication overdoses — accounted for the largest fraction of claims, the most severe patient harm, and the highest total of penalty payouts. Diagnosis-related payments amounted to $38.8 billion between 1986 and 2010, they found.

Researchers estimate the number of patients suffering misdiagnosis-related, potentially preventable, significant permanent injury or death annually in the United States ranges from 80,000 to 160,000.

According to the information from the publication more than 12,000,000 Americans are affected by diagnostic errors each year. Every error may lead to serious injury, disability or even death.

  • Access to high quality primary care (limited access due to lack of money, remoteness, illiteracy, travel constraints or a limited number of health care facilities)
  • Availability of health care professionals and specialists (lack of sufficient, competent health care professionals, for example, due to lack of training, outward migration or a poor employment situation. Specialty expertise may not exist or may be limited in number or quality);
  • Teamwork (poor teamwork, lack of learning and feedback when errors occur);
  • Availability of diagnostic tests (diagnostic tests limited in scope, availability or quality);
  • Communication (little or no sharing of medical information);
  • Care coordination (consultations delayed or test results lost or a lack of health records documenting care);
  • Follow-up (limited follow-up reduces the ability for diagnostic impressions to evolve);
  • Affordability of care (care unaffordable or compromises other basic needs such as food or housing);
  • Training of health care providers (training is suboptimal, in particular lack of training for clinical reasoning; certification and licensure requirements are deficient);
  • Availability of health informatics resources (health informatics resources, including internet access, may not be available, especially in remote areas; unaffordable subscription or download fees for medical information);
  • Culture (some cultures may be punitive, which discourage sharing and inhibit learning; physician-centric systems limit the value of the team. Patients may feel it is more appropriate to be passive care recipients);
  • Human factors and cognitive issues (the work environment and systems may be subject to distractions, interruptions and a lack of organization of information).
Conceptual model of missed opportunities in diagnosis. Source: WHO

The solution is almost here

Neural networks can already diagnose diseases on par with the best doctors by analyzing patients’ test results, X-ray images, etc. The best neural networks are developed not by large corporations, but by independent labs and research centers. Thousands of new participants are going join the neural network development market.

Skychain will combine their efforts by providing three useful services:

A single marketplace that will allow millions of healthcare professionals around the world access neural networks, paying the developers a fee for each use of their neural networks;

A neural network training platform that will allow the developers to train their neural networks for a fee on many datasets prepared by independent data providers, and get very deep and very accurate neural networks;

A distributed supercomputer that will host and run more complex and more accurate neural networks than any local data centers or servers could ever do.

Skychain will help medical data providers in the following ways:

It will provide a data marketplace to allow many data providers provide their datasets as a service for neural network training.

The neural networks trained on the medical data providers’ datasets will only be used within the Skychain system to guarantee that the data providers will get their share of the profits generated by training neural networks on their datasets;

Preparing medical datasets for neural network training will become a profitable business for the owners of such big data.

The Skychain project cares about the intellectual property of neural network developers and data providers. Both the trained neural networks and the datasets used for their training will be protected from any third parties.

Skychain will also protect the patients’ private data by sending them in a de-identified form and using end-to-end encryption.

Join Skychain on social media: Twitter, Facebook, Telegram

Egor Chertov, Skychain team

Blockchain infrastructure aimed to host, train and use artificial intelligence (AI) in healthcare. Our website: