3,772 Medical Incidents Reported in Ten and a Half Years, About 356 a Year — the Design Estimate Was 1,300 to 2,000, and the Clause Requiring a Review Within Two Years Has Sat Expired for a Decade

On 7 August 2026, NHK reported an announcement by Kyoto University Hospital. During an operation the previous month on a woman in her fifties for a brain tumour, surgeons removed not the tumour but part of the healthy brain governing functions including respiration. The patient now requires a ventilator.

The machinery for cases of this kind is the medical incident investigation system. It came into force in October 2015 under the sixth amendment to the Medical Care Act, promulgated in June 2014[1].

The cumulative totals are published. As of the end of April 2026, reported medical incidents total 3,772[2]. Over ten years and seven months, that averages about 356 a year.

The head of the institution decides whether to report

The scope is set in statute. It covers "a death or stillbirth arising from, or suspected to arise from, medical care provided by a healthcare worker employed at the institution, which the head of that institution did not anticipate"[3].

Note the subject of that sentence. Whether the death was anticipated is decided by the head of the institution where it happened. Not by a third party.

A further note is attached: whether an error occurred is not part of the test[4]. The line is drawn on anticipation, not on fault.

Anticipation can be established three ways

The ministerial ordinance lists three ways a death counts as anticipated. Where the patient or family was told before care was given that the death was anticipated; where it was recorded in the medical record or another document before care was given; or where, after hearing from the healthcare workers involved and taking the view of the medical safety committee, the head recognises that it was anticipated before care was given[5]. An incident falls inside the system only where the head finds that none of the three applies.

The third is settled after the fact. Hear the staff, take the committee's view, and the head may find the death was anticipated. Discretion belonging to the party involved sits at the entrance to reporting.

The notice adds a limit. The explanation and the record must concern the possibility of this death for this patient in light of their clinical course, not a general statement about the possibility of death[6]. Generalities are not to serve as cover.

The estimate was 1,300 to 2,000 a year

What volume was expected when the system was built? The ministry's own material says.

The estimate at the design stage was 1,300 to 2,000 a year[7]. Eight months after commencement the record stood at 251 incidents reported, 1,250 consultations, 78 investigation results filed and 2 Centre investigations requested[7].

In the same document the ministry qualifies its own estimate. The estimate was built before the system's scope was settled, from death counts under an existing reporting scheme covering university hospitals and National Hospital Organization facilities. It assumes large hospitals with a high share of high-risk patients, and it includes deaths other than those "the head did not anticipate," so the estimated figure runs high[8].

Ten years on, the record averages about 356 a year, 27.4% of the lower bound of 1,300. Whether the estimate was too high or the reporting is too low cannot be settled from these figures alone. It has stayed unsettled for a decade.

Consultations total 20,213, half of them from providers

Contact with the system runs far higher. Consultations total 20,213: 9,446 from healthcare providers, 9,501 from bereaved families and others, and 1,266 other or unknown[9].

About half come from providers. Consultations from families and from institutions stand at almost the same number. The desk is both a place families turn to and a place institutions call when a judgement is unclear.

The second use has its own record. Where a provider sought advice on whether an incident fell within the system, the Centre convened a panel and advised 691 times in total[9]. That is 18.3% of the number of reports.

The criminal route sits outside the system

The ministry's material sets a second route alongside the diagram of the system.

Notification under Article 21 of the Medical Practitioners Act, made on the judgement of the doctor who examines the body, leads to police investigation, a prosecutor's decision, referral, and then charge or no charge[10].

Reporting under the system and notification under Article 21 run as separate routes. One exists for prevention, the other for investigation. A single death may travel both, or neither.

The system side states its purpose. The purpose of this system is to secure medical safety; it does not exist to pursue individual responsibility[11].

There is a proviso on how interviews are handled

The two routes surface again in how investigations run. Interview records for the healthcare worker involved are to be handled as internal material and not disclosed, except where there is legal compulsion, and the interviewee is to be told so[11].

The exception carries the weight. Where there is legal compulsion, the record is disclosed. Nothing in the wording of the notice guarantees that what someone says honestly will stay out of an investigation or a lawsuit.

The same notice states that the healthcare worker involved is not to be excluded from the investigation, and that anonymity is to be preserved as far as possible during it[11]. Investigating the person while preserving their anonymity is left to practice to reconcile.

The statute required a review within two years

This is where the system hangs.

The supplementary provisions required that, within two years of promulgation and having regard to how investigations were being carried out, consideration be given to notification under Article 21 of the Medical Practitioners Act and reporting to the Centre, to the form of medical incident investigation, and to the form of the Centre itself, and that legislative or other necessary measures be taken[12].

The deadline was 24 June 2016. Before it fell, the ministry wrote this. On the form of the system, opinions vary among those concerned on matters including Article 21 of the Medical Practitioners Act and the relationship between medical practice and criminal responsibility, and at present no amendment can be made by the deadline of 24 June 2016 set in the supplementary provisions[8].

A matter the statute directed be considered and acted on within two years went unprocessed by the deadline. The reason named is disagreement over the relationship with criminal responsibility.

What went in instead were operational changes

In place of amendment, five operational improvements were set out. Giving formal standing to liaison councils of support organisations, clarifying that heads of institutions must have arrangements to capture every in-house death, clarifying that the Centre relays the content of a family's consultation to the provider at the family's request, expanding training for support organisations and providers, and clarifying that the Centre may query the content of an in-house investigation report with the provider's consent[8].

The third has a record. The Centre relayed the content of a consultation to a provider at a family's request 253 times in total[9]. That is 2.7% of the 9,501 consultations from families.

The improvements went in; the matter the statute directed be considered stayed where it was. Four of the five concern arrangements and support on the provider side. None touches the relationship with criminal responsibility.

How often families have to ask for an investigation

The Centre investigation figures are worth setting out. Centre investigations were requested 329 times: 261 by bereaved families and 68 by providers. Reports were issued 229 times[9].

Requests come to 8.7% of the 3,772 reports, and 69.6% of requests produced a report. 79.3% of requests came from families. The mechanism is used by the side that could not accept the result of the in-house investigation.

In-house investigation results filed total 3,372[9], 89.4% of reported incidents.

95.1% of reports come from hospitals

The origin of reports is lopsided too. Of the 3,772 reports, 3,586 came from hospitals, 185 from clinics and 1 from a midwifery centre[9]. That is 95.1%, 4.9% and one case.

The system covers hospitals, clinics and midwifery centres alike. One report from a midwifery centre and 185 from clinics in ten and a half years. That does not track the number of facilities. Whether an institution can assemble an investigation in-house appears to bear on whether anything is reported at all.

What has to be counted before the next decision

What is published is counts of reports, consultations and investigations. Two things needed for a judgement are missing.

The first is the number of deaths not reported. The scope is deaths "the head did not anticipate," and a death the head finds was anticipated falls outside. The number found to fall outside is tabulated nowhere. Whether the estimate of 1,300 to 2,000 was too high or the reporting too low cannot be settled without it.

The second is the number that travelled the criminal route. No published statistic matches notifications under Article 21, investigations and charges against reports under the system. Nobody knows how the two routes divide the same deaths, while the clause meant to reconcile them sits expired.

As the shared structure of programmes that fail to reach people（このサイトの記事） sets out, a system existing and a system being used are different things. Here the number of uses is known. The number of non-uses is not.

3.456 million consultations over ten years of support for the needy（このサイトの記事） showed the same shape. Those who entered are counted; those who did not are not. In a system whose entrance is left to the judgement of the party involved, that asymmetry becomes the limit on what can be evaluated.

Further Reading

• Iryo Jiko to Keiho (Medical Accidents and Criminal Law)（外部サイト、新しいタブで開きます） by Katsunori Kai (Seibundoh). A study of the relationship between medical practice and criminal responsibility from the side of criminal law. Useful for thinking about why the matter the statute directed be considered within two years has not moved in ten.

References

Outline of the Medical Incident Investigation System — Ministry of Health, Labour and Welfare, Medical Safety Promotion Office (2016). Ministry of Health, Labour and Welfare

Status Report on the Medical Incident Investigation System (April) — Japan Medical Safety Research Organization (Medical Incident Investigation and Support Centre) (2026). Japan Medical Safety Research Organization

On the Medical Incident Investigation System — Ministry of Health, Labour and Welfare (2026). Ministry of Health, Labour and Welfare

Medical Incident at Kyoto University Hospital: Part of a Healthy Brain Removed in Error During Tumour Surgery — NHK (2026). NHK NEWS WEB