Health care databases are not created in a vacuum. They represent patients—people—and their structure and content are therefore shaped by the cultural, economic, and political characteristics of the populations they represent.

It is essential, then, for researchers to understand the source of the data they’re studying to work with any database, as cultural norms, disease prevalence and incidence, and particularities in health systems can affect what researchers see in the data.  

“This is something that we struggle with [when studying] multinational databases in general,” says Soko Setoguchi, M.D., Dr.PH., associate professor of medicine at the RWJ Medical School and Center for Pharmacoepidemiology and Treatment Science, Institute for Health at Rutgers Biomedical and Health Sciences. “We look at the same things in different databases and from different countries, and we sometimes find striking differences between the countries, or among the countries.” 

Take, for example, a Japanese database: the Japanese Medical Data Center (JMDC Inc.). 

JMDC Inc., part of the Aetion Data Network, was founded in 2002 and is one of the largest commercial data sets in Japan. It represents about 13 million patients across two different databases—payer data and hospital data—and shows longitudinal patient views using data from hospitals, clinics, pharmacies, and dentists’ offices. Hundreds of studies have been published using JMDC Inc. data in therapeutic areas like diabetes, cardiovascular disease, and rare disease. 

Understanding this data first requires an understanding of the Japanese health system. Key differentiators include:

  • All Japanese citizens receive National Health Insurance (NHI). Either private or public, provided through their employer or through the government, everyone in Japan is insured. To care for its elderly population—Japan has the oldest population in the worldlong-term care insurance (LTCI) is offered to all citizens older than 65. Because Japanese employees change jobs infrequently, less health data are lost than in nations where employees change jobs or insurers more often. National insurance covers most medical services, with the exceptions of over-the-counter drugs, normal pregnancy and delivery care, vaccines—which are paid for by local governments—and “lifestyle” treatments like cosmetic surgery. JMDC Inc. data is sourced from private health insurance associations, provided mostly by employers, and therefore represents a younger population of working age people and their families. 
  • Japanese patients generally do not have one primary care physician. NHI allows citizens to visit any clinic or hospital at any time. Many visit those which are convenient to their homes or work. 
  • Annual check-ups are mandatory. Known as the kenko shindan, annual health checks are mandated by employers for all Japanese people over 40—part of a national focus on wellness known as Health Japan 21. The set of tests and checks administered can include blood work, a chest X-ray, height and weight measurements, and vision, hearing, urine, blood pressure, and obesity tests. People can also opt for an annual “Ningen Dock,” which is a day-long or overnight stay in a hospital in which patients receive a full health work-up, including an endoscopy, cancer screenings, X-rays, and other tests. People are given a letter grade between A and F on their health, and results are shared with employers. If a result is considered abnormal or of concern, employers discuss it with the employee (Japan’s Act on the Protection of Personal Information (APPI) allows for this exchange, where the U.S. equivalent, the Health Insurance Portability and Accountability Act (HIPAA), does not). 
  • In order to have a prescription refilled, a patient must see a doctor. In Japan, doctors cannot write and pharmacists cannot refill prescriptions. Each time a patient needs a new prescription to refill one, they must consult a doctor and prescriptions for new drugs are often only written for two weeks at a time. Patients with long-term or chronic conditions may receive prescriptions for up to a year for most medications except narcotics. 
  • Insurance claims are submitted monthly, rather than an encounter-by-encounter basis. Patients and health care facilities submit a monthly report of insurance claims per patient. This allows researchers to see which claims were submitted in which month, but not necessarily the order in which they were submitted. 
  • Inpatient and outpatient claims are paid for differently. Payment for inpatient care is processed through a uniquely Japanese system known as the diagnosis procedure combination (DPC). Inpatient DPC hospitals charge a flat rate for the inpatient stay, which is calculated by multiplying the rate by the length of the stay, plus additional costs for surgeries or other procedures. Outpatient care is fee-for-service. 
  • Japanese data use a Japan-specific coding system. These data can be mapped to ICD-10 codes, but there is no one-to-one translation for the country specific drug, procedure, and diagnosis codes. In this coding system, patients are identified by their first and last names, so if someone changes their name, for example, when a woman marries and changes her name, their data may not be traceable to them. Another differentiator of note for this data set: death data is not documented unless a patient dies in the hospital.
  • Japanese data is not linkable. In some US databases, data collected in hospitals via EHRs are linked with claims data from insurers. This practice is currently prohibited in Japan. 

Taking all of these and other factors into account, a researcher can then work with the JMDC Inc. database with a greater understanding of Japanese health care and the data it generates. 

Working with international databases
Working efficiently and effectively with real-world data from multinational data sets can be done with a few key best practices and tools. A validated, real-world evidence (RWE) platform that can work with data in a transparent and traceable way is one. When scientists have the expertise needed to work with data like the JMDC Inc., a platform that is data-fluent can help them integrate it and draw insights regardless of its structure or coding language, and without scrubbing it of the captured events that make it unique.

Of utmost importance, though, is a deep understanding of the data source that you plan on working with. 

“You need scientific expertise relevant to a given data source,” says Sebastian Schneeweiss, M.D., Sc.D., professor of medicine and epidemiology at Harvard Medical School and Brigham and Women’s Hospital. “There is not a single person who knows every database and all aspects of a given database. It is therefore important to have a mechanism in place through which you can reach out to the people who know best how a specific data source was generated and what each feature in the data really means.”