Determination of the data sources used for non-interventional observational studies in Japan
We used a binary approach to determine the data sources used for past non-interventional observational studies conducted in Japan.
In the first part, we conducted a PubMed search to find articles reporting observational studies conducted in Japan in the last 3 years; the following search settings were used: affiliation field, Japan; article type, observational study; publication date, December 1, 2017 to November 30, 2020; species, human; and language, English. We excluded articles without an abstract or without a structured abstract; articles reporting studies conducted multi-nationally or in a country other than Japan; and articles reporting non-clinical studies, interventional studies, or studies with healthy subjects/controls.
In the second part, we obtained a list of articles reporting observational studies using data from one or more of four large-scale Japanese EHR databases8–12: Japan Medical Data Center Claims Database (JMDC Claim), Medical Data Vision Database (MDV Database), National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB Japan), and Medical Information Database Network (MID-NET). JMDC Claim and MDV Database are the largest EHR databases in Japan1. NDB Japan and MID-NET are widely known databases in Japan that are provided by the Japanese governmental organization. These four databases all include EHR data generated by the Japanese Diagnosis Procedure Combination/Per-Diem Payment System (DPC; the Japanese medical payment framework); therefore, these databases are sometimes colloquially referred to as “DPC databases”. For example, NDB Japan is frequently called the “DPC database” because it is the most well-known database in Japan.
To obtain this information, we accessed the list of published articles available at the website associated with each of the four databases. For JMDC Claim and MDV Database, the publication lists were very long so we limited our search to the period January 2017 to June 2019. For NDB Japan and MID-NET, the publication lists only included articles published since 2018.
After compiling our list of articles, we reviewed each abstract for information about data source, target disease or therapeutic area, number of patients enrolled in the study, and study design. If the abstract did not include this information, we reviewed the full text of the article. If the full text was not available or lacked enough information, we excluded the article. Classification of each article was conducted using the following criteria:
Data source: Articles were classified into four types based on the source data used: (1) data stored at a medical institution, (2) data collected from several medical institutions, (3) data obtained from a disease registry or database, (4) data obtained from a large-scale integrated database. When the words “single-center study” were included in the abstract or full text, the article was classified as type (1); when an article included the words “in a hospital”, “in our hospital”, “in our institution”, “at XXXX [institution name]”, “in a unit”, “in a center”, or “in a department”, we assumed that this indicated a single-center study, the article was classified as type (1). Similarly, when the abstract or full text included the words “multi-center study” or “in hospitals”, the article was classified as type (2); when they included “registry” or “study database” it was classified as type (3); and when they included “large database” it was classified as type (4). Articles with the words “nationwide database” were also classified as type (3) or (4) after we carefully checked the actual name of the database, if available, in the rest of the abstract and full text. Because post-marketing surveillance (PMS) studies are a kind of multi-center study, these articles should also be classified as (2). However, in Japan, PMS studies are conducted under Good Post-marketing Study Practice regulations13, whereas observational studies are conducted under the ethical guidelines of the Japan Ministry of Education, Culture, Sports, Science and Technology14; thus, PMS studies were considered separately from the other multi-center studies.
Target disease or therapeutic area: Articles were classified by using the following 19 classes with reference to the 10th revision of the International Statistical Classification of Disease and Related Health Problems15: (1) infectious and parasitic diseases other than coronavirus disease 2019 (COVID-19), (2) COVID-19, (3) cancer and neoplasm, (4) diseases of blood and blood forming organs, (5) endocrine, nutritional and metabolic diseases, (6) diabetes, (7) mental disorder, (8) nervous system, (9) disease of the eye and adnexa, (10) disease of the ear and mastoid process, (11) circulatory system, (12) respiratory system, (13) digestive system, (14) Disease of the skin and subcutaneous tissue, (15) disease of the musculoskeletal system and connective tissue, (16) disease of the genitourinary system, (17) pregnancy, childbirth, and perinatal, (18) injury or other consequences of eternal causes, (19) others, including surgery, transplantation, hemodialysis, dental, and pain.
Number of patients enrolled in study
The number of patients included in the final analysis was obtained by reviewing the abstract or full article text.
Study design
Studies were classified as either prospective or retrospective, depending on which word was used in the abstract or full article text.