Clinical Records and Medical Research

Under Article 24 of Japan's Medical Practitioners Act (医師法), medical records must be retained for 5 years (for insured care, the Rules on Medical Care Benefits set the retention period at 5 years from the date of completion). 2 years is the retention period for other records related to care at a hospital (nursing records, X-ray films, etc.) under the Medical Care Act (医療法) and its enforcement regulations; 3 years is the retention period for ledgers, etc. under the Rules on Medical Care Benefits; there is no 10-year requirement.

Under the Medical Care Act (医療法) and its enforcement regulations, a hospital's records related to care (the hospital day book, prescriptions, operation records, nursing records, test findings, X-ray films, etc.) must be retained for 2 years. 5 years is the retention period for medical records under the Medical Practitioners Act, and the differing legal basis and period between these records and medical records is a point often tested. There is no 1-year or 10-year requirement.

O stands for Objective information, recording facts observed or measured by the healthcare provider, such as examination findings and test results. Subjective information such as the patient's complaints is S (Subjective), evaluation and consideration is A (Assessment), and the plan is P (Plan). SOAP is the standard format for writing progress notes in the POMR.

The POMR is a method that organizes records around each of the patient's medical problems, so the statement that records are organized by each department's revenue is incorrect. Based on the concept of the POS, it consists of the database (basic data such as the patient's basic information), problem list, initial plan, and progress notes, and the progress notes are written in SOAP format.

The discharge summary is a record in which the physician summarizes the course of care, treatment content, and outcome during hospitalization; it is used for continuity of care in outpatient clinics and referral institutions and for audits and statistics, and is completed after a countersignature (approval). It is not a document the patient prepares, a dispensing ledger, or a financial document.

Auditing medical records includes a quantitative audit, which checks whether the required documents, entries, and signatures are complete, and a qualitative audit, which checks whether the content is appropriate in terms of quality of care. Auditing is also performed in-house, such as by the health information management department. A qualitative audit is not about counting the number of records, and records are audited regardless of whether they are on paper or electronic.

Outcome is a category representing the result of care, including cure, improvement, remission, no change, worsening, death, and discontinuation. The chief complaint is the main symptom the patient reports at the visit; it is information recorded at the start of care and is not an outcome category.

The medication notebook is a record of dispensed medication history that the patient carries with them, and it is one of the health records — along with the maternal and child health handbook and health checkup records — issued to the patient. The order sheet, operation record, and anesthesia record are in-house medical records created and kept by the healthcare institution, not records kept by the patient.

Medical records must have authenticity — it must be clear who recorded or corrected what and when, and tampering must be prevented. For paper records, it is appropriate to correct with a method such as a double line that keeps the original entry legible, and to make clear who corrected it and when. Erasing with correction fluid, discarding the page, or overwriting anonymously in pencil destroys the original entry or the accountability for it and undermines authenticity.

The ICD is the WHO's international classification of diseases, forming the basis for international comparison of cause-of-death and disease statistics; Japan's statistics are also currently based on ICD-10. The surgery codes in the medical fee schedule are K codes, the barcodes for distributing drugs and other items are GS1 codes, and the domestic standard code for clinical laboratory test items is JLAC10 — none of which describes the ICD.

Japan's cause-of-death and disease statistics are currently compiled using a statistical classification based on ICD-10, and the WHO adopted its revision, ICD-11, in 2019 (effective 2022; Japan is still preparing for domestic application). ICD-11 is a newer revision than ICD-10. The ICD is the WHO's international classification, not a domestic standard developed by a Japanese academic society.

The ICD-10-based Standard Disease Name Master is a standard master that maps disease names to ICD-10 codes and the like; it is maintained by the Medical Information System Development Center (MEDIS-DC) and is widely used in electronic medical records and electronic claims processing systems. It is not a master for surgical procedures, imaging conditions, or the logistics of medical materials.

The DPC code has 14 digits, with the first 6 digits representing the disease name (the first 2 digits are the major diagnostic category, MDC, and the following 4 digits are the sub-classification), and the remaining digits represent surgery, procedures, and the like. DPC is a diagnosis procedure combination classification for acute inpatient care, not a classification for outpatient appointment management. The code is determined from the diagnosis and care provided, not generated from the patient's name. DPC/PDPS is a per-day bundled payment system.

In the medical fee schedule, surgery is represented by K codes (the classification numbers starting with K). J codes, in the same fee schedule, represent procedures. The HOT code is the code of MEDIS-DC's standard drug master, and JLAC10 is the standard code for clinical laboratory test items; neither is a surgery code.

JLAC10 is the standard code for clinical laboratory test items, used for sharing and aggregating test results between facilities (LOINC is the equivalent international standard). ICD-O-3 classifies tumors, GS1-128 is a barcode symbology for drugs and medical materials, and the K code is the surgery code in the medical fee schedule; none is a clinical laboratory test item code.

An RCT is an interventional study that randomly assigns participants to an intervention group and a control group to compare effects; random allocation reduces bias (confounding) in the background characteristics between groups. Examining at a single point in time describes a cross-sectional study, tracing past exposure from disease status describes a case-control study, and making no intervention describes an observational study.

A case-control study is a retrospective observational study that first selects a group with the disease (cases) and a group without it (controls), then compares their past exposure; it uses the odds ratio as its measure and is suited to studying rare diseases. Random allocation is used in RCTs, following forward in time is a cohort study, and using population-level data is an ecological study.

A cohort study is a prospective observational study that follows a group exposed to a factor and a group not exposed forward in time and compares the occurrence (incidence) of disease, allowing the relative risk to be calculated directly. Not following participants describes a cross-sectional study, assigning treatment describes an RCT (interventional study), and a cohort study does set up comparison groups — the exposed and unexposed groups.

Evidence level is generally highest for a systematic review/meta-analysis that systematically integrates multiple RCTs, followed in descending order by individual RCTs, cohort studies, case-control studies, case reports, and expert opinion. An individual expert's opinion and a single case report rank near the bottom of the hierarchy.

A confounding factor is a third factor that is associated with both the exposure (factor) and the outcome (disease) being studied, creating an apparent association that does not actually exist, or distorting a real association. Random allocation in an RCT and stratified analysis are ways to control for its effect. Error from measurement equipment is measurement bias, and dropping out partway through is a loss-to-follow-up problem, neither of which is the definition of confounding.

The Declaration of Helsinki, adopted by the World Medical Association, sets out the ethical principles for medical research involving human subjects and also underlies Japan's ethical guidelines. In principle, a research plan must be reviewed and approved by an ethics review committee (IRB), and a conflict of interest must be disclosed and managed, not concealed. Falsification and fabrication are research misconduct regardless of the conclusion.

Opt-out is a method that notifies or publicly discloses information about the research and guarantees participants an opportunity to refuse participation (use of their information), then uses existing information for the research if no refusal is made; it is permitted in certain cases, such as observational research. Obtaining individual consent beforehand is opt-in (informed consent), and methods that give no notice at all, or notify only after the fact, do not meet the requirements of opt-out.

A phase I trial is, in principle, a stage conducted on a small number of healthy adults to examine safety and pharmacokinetics (absorption, excretion, and the like in the body). Verifying efficacy in a large number of patients is a phase III trial, and the post-approval survey is phase IV. A toxicity study on animals is a nonclinical study conducted before entering a clinical trial (a study on human subjects).

The median is the middle value when the data are arranged in ascending order; with 5 data points, it is the 3rd value, 5 days. 23.8 days is the mean of this data, pulled upward by the outlier of 100 days. With data containing an outlier like this, the median better represents the typical value than the mean.

In testing, the null hypothesis ("there is no difference") is set up, and the probability (p-value) of obtaining the observed result or a more extreme one under the null hypothesis is calculated; if it is below the significance level (conventionally 0.05), the null hypothesis is rejected and the result is judged significant. The p-value is not itself the probability that the treatment is effective, a smaller p-value indicates greater significance, and the null hypothesis is set up as "there is no difference," not "there is a difference."

Sensitivity is the proportion of people with the disease who test positive: 90÷100=90%. Note that specificity is 810÷900=90%, and the positive predictive value is 90÷(90+90)=50%; do not mistake 50% for the sensitivity. 10% corresponds to values such as the false-negative rate or prevalence, and 81% results from a miscalculation.

Specificity is the proportion of people without the disease who test negative: 720÷800=90%. 80% is this test's sensitivity (160÷200), approximately 67% is the positive predictive value (160÷240), and approximately 95% is the negative predictive value (720÷760); none of these is the specificity. It is important to check which numerator and denominator apply using a 2×2 table.

Of the 100 people with the disease, 100×0.8=80 test positive (true positives); of the 900 people without the disease, 900×0.1=90 test positive (false positives). Positive predictive value = 80÷(80+90) ≈ 47%. Even with 80% sensitivity and 90% specificity, when prevalence is low, false positives increase and the predictive value drops considerably. Approximately 98% is close to the negative predictive value (810÷830).

The ROC curve is drawn by varying the test's decision criterion (cutoff value) and plotting sensitivity on the vertical axis and the false positive rate (1 − specificity) on the horizontal axis; the larger the area under the curve (AUC), the higher the test's discriminative ability. A description with the axes reversed, or claiming a smaller AUC is better, is incorrect, and the curve does not represent the relationship between cost and time.

The NDB (National Database) is a national database, established under the Act on Assurance of Medical Care for Elderly People, that collects claims information and specific health checkup information and is used for analyzing medical costs and for research. The NCD is a clinical database of surgical cases and the like, centered on surgical societies. Hospitals have a legal obligation to report to the National Cancer Registry under the Cancer Registration Promotion Act. DPC data is data on acute inpatient care from DPC-participating hospitals.