Deaths by year this week: how archives review reports

For a celebrity memorial archive, most of the work on an annual record is not adding entries. It is deciding what to do with death reports that remain unresolved, and that review happens weekly rather than yearly.

Set this week's reports beside the year's record and the gap shows up fast. One is a pile of claims, the other a file of entries with sources behind them. Nobody is named here and no death is reported. What follows is about the queue behind the page.

What to take away

How weekly reports become annual entries

A weekly item is a name, a date and a claim, usually carried by one outlet. An annual entry is a name, a date and a source that survived a second look.

A desk can assess a new report's place in the annual file using four operational questions: how fast the record appears, who stands behind the name, whether the identity is settled, and what it takes to reach the record.

The pattern runs one way. The fastest layer, the weekly file of news reports, is the layer with nothing behind it. The slowest layers carry the authority.

CDC WONDER and the National Center for Health Statistics (NCHS) final annual mortality file carry no names in public output, so they can confirm a count but never a person. A compiler uses them to check that a year's total is plausible, not to settle one entry.

The state registries and the UK General Register Office hold the document that settles a name. In England and Wales, a death is typically registered within five days, whatever the newspapers printed or skipped.

The Social Security Administration's Death Master File and the National Center for Health Statistics' National Death Index sit in between. Both are name-level, and both are closed to the casual reader: the full Death Master File is limited to certified users, and the National Death Index requires an application and a fee.

Neither of those is a weekly tool. That is what the comparison shows: nothing that settles a name moves at the speed of a weekly report.

The weekly file is the only layer a desk can act on in seven days, and it is the least reliable one. Keeping that file honest until a slower record catches up is the weekly job.

What sits in the queue

A report reaches a desk and cannot be published. That happens for a small number of reasons, including an unclear identity, a single-source claim, or a source with direct knowledge who has not responded.

A claim repeated by several outlets can look corroborated without being independently verified.

Standing newsroom commitments to verification, of the kind gathered in the Associated Press statement of news values and principles, turn on whether a source had direct knowledge rather than on how many places carried it.

Each problem has a different resolution. Treating them as one bucket called "unconfirmed" is the mistake. A single-source claim needs independent confirmation. An ambiguous identity needs a distinguishing detail. A source that has not responded needs another documented contact or a decision to hold or drop the item.

Sorting the queue by why an item is stuck turns a pile into a work list.

The weekly pass

Once a week, every item in the queue gets looked at, and each look ends in one of four outcomes.

The four outcomes are publish, chase, hold, and drop. Publish when verification is sufficient; chase when a follow-up could resolve the issue; hold when a response or record is pending; and drop when the claim cannot be substantiated or should not be published.

The discipline is that an item cannot stay in the same state twice without somebody writing a sentence about it. That requirement stops a queue from turning into an attic.

Aging, and why it matters more than volume

As hypothetical examples, a queue of two hundred items where nothing is older than three weeks is healthy. A queue of thirty where the oldest is fourteen months is not, and the second one looks better on any count of size.

The number to watch is age, not volume. Old items are dangerous in one specific way: someone who was not there when they arrived re-reads them and lacks the context that made them doubtful.

That reader sees a plausible claim that apparently survived a long time. Survival is not corroboration, and an item in the queue for a year has failed to be confirmed for a year.

The other reason to watch age is that the routes to confirmation close. People who would have answered a question in the first fortnight are harder to reach a year on. Institutional records catch up on their own timetable, which is useful for the older cases.

In England and Wales, a local registrar records a death under the Births and Deaths Registration Act 1953, and the General Register Office maintains the national record. The linked local guidance on registering a death describes the registration steps.

What a dropped item leaves behind

This matters for a reason that only becomes obvious the second time. The same report will come back. Somebody will send it again, or it will resurface somewhere and look new.

A desk with a closed file can answer in a minute that this was looked at, that these were the routes tried, and that it did not stand up. A desk without one starts from scratch and reaches a different answer. That is how a claim that failed twice gets published on the third pass.

Where this sits

The queue feeds the annual file deaths by year. The same items pass through shorter windows: deaths this week and deaths by month.

Same-day handling of an unconfirmed report is a separate problem, addressed in deaths today. If a queue item is stuck on which of two people is meant, the identity methods in actors who died can help resolve identity in a celebrity memorial archive.

Common questions

How long should something stay in the queue?

There is no universal limit. Set an internal review period, revisit each item weekly, and document any extension and why it is warranted. An open-ended queue leaves reports unresolved without a clear decision.

Is a report from a reputable outlet enough on its own?

Not by itself, because reputable outlets copy each other. What matters is whether the outlet had direct knowledge or was repeating a source you can also see.

Can the year's record settle a name this week's file could not?

An annual record can help settle a name when a state vital records office or an authorized National Death Index search confirms the identity. Those records do not answer a weekly report immediately, and access to them may be restricted.

What if the family has asked for nothing to be published?

Then nothing is published. That is a decision rather than a pending item, and it belongs in the closed file with the reason recorded.

Should the queue be public?

No. A list of people somebody has heard might have died is exactly the kind of document that causes harm if it leaks, and it contains claims that have failed a check.

Does a dropped item ever come back?

Yes, when new evidence arrives. The closed file makes that cheap, because the next person can see what was already tried instead of repeating it.