By Dr Tom Foley.
Professor Tanaka is Professor Emeritus in the School of Medicine at Tokyo Medical and Dental University. He is also Special Advisor to the Executive Director of the Tohoku Medical Megabank Organisation.
The Megabank Project has built a database that integrates genomic as well as lifestyle data for a cohort that will reach 150,000 individuals, from the Miyagi Prefecture, by 2017. The project has been funded, with $500m, to follow up participants for at least 10 years.
8000 participants of the cohort will eventually have their entire genome sequenced (3000 participants’ genomes have already been sequenced), while the rest have had arrays measured, with other sections of their genome being imputed – a more cost effective approach.
Each participant completed a 31 page lifestyle/environmental exposure questionnaire and has had a range of blood tests taken on induction. This yielded 6,000 items of information per participant. The questionnaires will be measured repeatedly (using internet [e-epidemiology]) every year, during the follow-up investigation.
All individuals were believed to be healthy on induction and already 10% have developed chronic diseases. Initially, patients self reported diagnoses, but the project is now moving to verify these with healthcare providers.
Now that the team have almost completed the cohort recruitment, they have begun to develop prototype predictive models, to identify an individual’s risk of developing a chronic disease, based on their genomic and lifestyle factors. The team have found that 6,000 items of lifestyle information and the individual’s full genomic information was too
much to analyse, so they have narrowed their focus to reduced number (about several hundred) of lifestyle/environmental factors and DNA ‘snippets’. It then becomes feasible to calculate the probability that a given disease will occur in an individual.
The team will require a longer follow up period, before these models can be fully utilised, but it is envisaged that they will eventually provide a website for participants to learn about their individual risks. The website will show the current risk of developing the disease, but will also offer lifestyle advice and show how, if followed, it would impact on their risk.
It appears that current diagnostic classification systems, such as ICD, do not adequately represent the true nature of diseases. For example, the team have identified several completely different mechanisms of action in different cases that have been classified under the same disease. The team’s work might eventually be extended to gain insights about which types of patients have responded well to treatments in the past. New patients could be fitted to these profiles, enabling an advanced form of decision support. This would present a number of challenges:
- Constructing patient profiles would be difficult, probably relying on stratification using correlation networks.
- Current Japanese EHR implementations lack the interoperability to provide sufficient routine data.
The data is deidentified and is currently available to researchers, working within the centre (a safe haven) and a small number of collaborating centres that are connected via VPN. Processing is only permitted on the centre’s supercomputer and only aggregated data is released publicly.
Beyond the Megabank project, Professor Tanaka discussed more general attempts to build Learning Health Systems in Japan. A recurring challenge has been to extract data from EHRs. Most large hospitals in Japan have implemented EHRs. ~90% of the market is dominated by, Fujitsu, Hitachi and Toshiba, with ~10% being held by IBM. EPIC, Cerner and Allscripts, were not thought to have a significant presence in the country. EHRs were thought to be popular
The Ministry of Health define the Medical Information Exchange Standards, that ensure some level of interoperability, however, these standards only cover basic data, such as, diagnosis and lab results. This has severely limited the potential for comparative effectiveness research, such as retrospective observational studies.
Attempts have been made to create regional health networks, however, small hospitals and family practices still use largely paper-based records.
Provider benchmarking is commonplace in Japan. Traditionally, this has been focused on financial metrics, but has expanded to include measures such as, diagnosis, treatments, admission rates, length of stay, procedure types and success rates, although pure quality or outcome measures are still rare. Standardised outcomes measurement is uncommon.
Surveillance systems are beginning to develop. The Japanese Institute of Infectious Diseases has built a system that can track outbreaks in near real time. Meanwhile, the Japanese Sentinel project aims to emulate the US FDA Mini Sentinel Project, by monitoring adverse drug reactions, automatically pulling data from EHRs in 10 university hospitals. This project is still under construction and results are awaited.