While countries such as the US have just a handful of software suppliers for handling medical data, Chinese hospitals are running the software from around 4,000 different companies. Even within a hospital, there is no consistency and a patient can have up to 40 different ID numbers from various departments. The government is hoping to establish three large national health databases by 2020, including medical notes.
But how are they going to make sense of the vast arrays of different types of data? Yidu Cloud (医渡云) has already processed the data of over 300 million patients by setting up cloud services in participating hospitals or among groups of hospitals. It works with medical research institutes to form partnerships with hospitals to use data sets for researching diseases and treatments.
China’s medical data
Hospitals are the frontline of healthcare in China. Patients head to hospitals for all medical care rather than visiting family doctors for referrals. There is widespread mistrust of hospitals and health professionals, all of which are perceived as trying to make as much money as possible from patients and their families for poor quality care. This results in more patients heading to larger, better-regarded hospitals which are the biggest in the world.
“Each of the hospitals we work with is huge. Each of the hospitals has 50 to 200 systems, with up to 20,000 outpatients per day,” Yidu Cloud founder Gong Rujing told TechNode, “So it’s extremely difficult to make the data from unusable to usable. Some of the data might have 40 different IDs for the same patient in different systems.”
92% of Chinese health professionals believe it is important that the country’s healthcare system be integrated, according to the Future Health Index 2017, operated by Philips. The current setup means patient data can be collected as per an individual doctor’s preferences with no standards even within a hospital, according to Yan Jun, Yidu Cloud’s Chief AI Scientist. This means data held by hospitals is difficult for hospitals themselves to understand and of no use to researchers needing data for drug and treatment development.
Clouds gather over hospitals
Yidu Cloud has invested $100 million in setting up data collection and analysis. “We need to deal with billions of data records. So only processing the data by human being manually is almost impossible. We need a machine learning approach to deal with the data. The first technical part is called ‘named-entity recognition,’ or NER,” said Yan Jun.
To do this, the company sets up a private cloud system within a hospital, or provides access to a joint cloud. Servers are set up in the hospital, not within Yidu. Then “the hospital authorizes us to get their data, they provide channels for us to access the data, or they just upload the data to our cloud,” explained Yan.
The company then applies AI algorithms to the various data formats to interpret it. This involves natural language processing (NLP) to understand the paragraphs of data on patients. But individual patient identity and records are anonymous.
Faster, cheaper, better: How China and AI are helping pharmaceutical development
Data security is vital both for hardware, staff training, and compliance. “We have to work with the Ministry of State Security to look at each of the private clouds, to make sure they are secured. We actually apply a way stricter law than HIPAA,” said Gong, referring to the Health Insurance Portability and Accountability Act–US legislation which governs the handling and transfer of data between hospitals and the patients’ insurance companies. “[In the US] they understand that under each different scenario you go through a certain identified compliance process. It’s very clear, there’s HIPAA law. But in China the policy is still being set up.”
For this reason, Yidu Cloud is already trying to be compliant above and beyond what HIPAA requires in the US while China formulates its own standards.
Medical research and the business of data processing
With the data processed and tagged, it is then useable in medical research. The hospitals pay Yidu Cloud to process their data and Yidu then finds research partners and brokers deals to connect the data providers–the hospitals–and these paying clients.
Yidu Cloud has established “strategic cooperation relations” with over 700 medical institutions in China and more than 100 top-class hospitals. The data, collected from over 300 million patients covering more than 30 types of critical disease, has led to the building of over 3,000 disease models. According to the company, its data platform boasts the largest quantity of data processed, the highest level of data integrity, and most transparent development process in China.
Getting to this stage took almost three years of work on the platform with no revenue before June 2017. “That’s very rare in China,” said founder Gong Rujing, “We were just investing in technology and 500 top scientists. We invested almost $100 million into our system to do it at a US standard. That’s difficult in a capital sense. But it’s even more difficult in terms of retaining the best talent. Because they can do an easier job like going to Toutiao, BAT, or doing fintech rather than trying to understand a huge paragraph of NLP and EMRs and medical problems and solving problems like this.”
The company expects to break even in the next couple of years. As well proving a viable business, processing medical data to make it useful for research is speeding up drug development. The real-world, unbiased data (data to which researchers can apply their own parameters) can be used to prove efficacy and safety which will lead to cheaper, faster drugs and treatment development.
Every cloud has a green lining
The data processing and collection systems being installed are leading to several positive side effects. The term “green healthcare” is used in China to sum up positive, efficient practices, free of fraud. Yidu Cloud claims their system saves time and money by minimizing the number of rewrites and repeat entries.
Standards are also an issue. Different hospitals, department, and staff have their own particular styles and standards of note taking. “All over the world, governments and academic organizations want to push or publish standards, and make everyone follow the standards. But in the clinical scenario, I don’t see many hospitals or doctors following such standards,” said AI scientist Yan Jun.
Yidu’s AI system means standards are no longer so necessary. According to Yan, “even though we didn’t publish any standards, after our processing, all the data from all the hospitals will follow the same standards.”
Depending on the compliance arrangements, the data can be used for international studies. Yidu Cloud is already looking beyond China for opportunities and is speaking to the UK’s National Health Service. Although it has a near monopoly on health services in the UK, the NHS has never managed to establish a centralized electronic medical notes system, despite spending over £10 billion on the project. Perhaps a Chinese data processor can help make sense of the situation.