Medical Examiner Case Management Problems: The 5 That Cost Real Money, and How to Avoid Them

Why does scoping the case as a linear workflow happen so often?

The pain presents as triple data entry, so the fix gets written as a workflow tool. Intake, autopsy, certification, closed, with a status dropdown and a dashboard counting how many cases sit at each stage. It demonstrates well, it is cheap to quote, and every generic case management product on the market is shaped exactly like it.

The diagnosis is wrong. A death investigation does not advance through stages. Toxicology comes back in six weeks and reverses a preliminary opinion. An anthropology consult on skeletal remains takes months. A dental comparison identifies a decedent who has been a case number for two years, and at that moment the demographics, the certification and the next of kin notification all activate at once, retroactively. A case signed out as undetermined is reopened when police develop new information.

A status field cannot represent that. What the domain needs is parallel evidentiary threads, scene, autopsy, histology, toxicology, anthropology, odontology and identification, each with its own owner, its own clock and its own completion state, converging on a certification that itself has versions.

The cost of getting it wrong is that you find out after go live, with live cases in the system, and re-modelling a case is not a change request.

The scoping test takes one question. Ask the developer to show you, in their proposed schema, what the certification said before an amendment and the date it changed. If the answer involves updating a field, stop the conversation there. That is the single most important design decision in this domain and it is visible before anybody writes code.

What goes wrong when you migrate an Access database and twenty years of photographs?

The second failure is quieter and it is genuinely dangerous. The legacy record is usually a Microsoft Access database built by someone who left in 2016, with free text fields, case numbering that changed format twice, and decedent names entered inconsistently. Alongside it sits a shared drive of scene and autopsy photographs organised by year and case number, mostly.

That word mostly is the whole migration. Photographs matched to cases by folder name will attach images of one decedent to another wherever the folder convention broke, and in this domain a mis-association is not a data quality issue, it is a case that cannot be used.

The third source nobody plans for is the paper log book at the cooler door, which in most offices is the most reliable record in the building. It knows who came in and when, and it frequently disagrees with the database.

Three rules keep this contained. Reconcile three ways before importing anything: the database, the photograph folders and the log book, and produce a written exception list of every case where they disagree. Never auto-associate a photograph whose folder cannot be matched with confidence, quarantine it for a human instead, and accept that a residual set will stay quarantined forever. And keep the legacy case number on every record without renumbering historical cases, because those numbers appear in court transcripts, police files and death certificates that cannot be reissued.

Why do the death registration and laboratory interfaces break after launch?

Two interfaces matter here and they fail in opposite ways.

The state electronic death registration system is a scheduling risk rather than an engineering one. Where a state exposes an interface the code is straightforward and the approval queue is not, and a project plan that assumes a technical timeline will slip against a bureaucratic one. Where no interface exists, the practical win is still large: eliminate retyping inside your own building and reduce the state step to reviewing prefilled content rather than authoring from scratch.

Reference laboratory results fail silently, which is worse. Results arrive as documents or structured files and have to be matched to the case. If matching is done on decedent name, a transposed date of birth or a name recorded differently by the laboratory produces an orphaned result. Nothing errors. The case simply keeps aging while everyone assumes the laboratory is slow, and the family calls at month five to tell you.

Three fixes, all cheap at build time. Match results on accession number and never on name, because the accession is the identifier both parties agreed at the point of submission. Route every unmatched result to an exception queue with a named person responsible for clearing it, not to a log file. And put a freshness alarm on each laboratory, so a provider that has sent nothing since Tuesday is visible on Wednesday rather than in a quarterly review, which is also how you build the evidence to renegotiate a contract.

What happens when records release and retention are not covered?

Your autopsy report may be subject to disclosure under your state's public records law while the photographs are exempt, while toxicology on a decedent whose death is under active investigation may be withheld entirely. The rules differ by state and sometimes by case status. Meanwhile a prosecutor needs the complete file and a civil litigant needs a subpoenaed subset.

The build failure is treating sensitivity as a property of a document. Once it lives at document level, every release means a person reading pages and blacking out lines, which is slow, inconsistent between staff, and impossible to prove afterwards.

The second failure is having no record of what was released. When a release decision is challenged, and eventually one will be, the office needs to show exactly what went out, to whom, on what date and under which exemption. An email with an attachment is not that record.

Retention is the third piece and it is architecture rather than configuration. Decades is a real requirement, and it means asking now what happens when the office changes cloud provider in year eight, whether the archive stays readable, and how integrity can be demonstrated for an image captured in 2026 and produced in court in 2044.

What belongs in scope from the start: sensitivity as a property of each data element so release packages assemble by rule, human review retained for narrative fields because nobody should promise otherwise, and an immutable release log.

Should you build custom or configure what you already own?

If you are a small coroner office handling a few hundred cases a year with one investigator and a contracted pathologist, do not build. A structured shared drive, consistent naming and the state registration system will genuinely hold that together, and a custom system you cannot afford to maintain is worse than a spreadsheet everybody understands. If your county is already funding a shared justice platform that intends to absorb you, do not build something the county plans to retire.

The signals that the arithmetic has flipped are concrete. You are accredited or pursuing accreditation, where turnaround statistics and case type distributions are effectively unmeetable by hand at volume. You are a state chief medical examiner office or a regional system covering multiple counties, where coordination between offices is the actual problem and no shared drive can do it. Your pending caseload contains cases nobody can account for. Or you have been asked for a complete file in litigation and assembling it took more than a day.

One observation about the empty vendor landscape. The absence of a dominant packaged product here is not a gap waiting to be filled, it is a signal that the requirements really are local, because jurisdiction, autopsy authorisation, records release and retention are all governed by state statute. Build accordingly, keep the state systems as the systems of record they are, and resist making the software the whole operation.

How do hidden costs get into the quote?

Four places, all visible at scoping if somebody asks.

Multi-county operation. Each county's law enforcement partner and each district attorney wants something slightly different, and that is parallel requirements rather than a configuration screen.

Imaging. Attaching postmortem computed tomography or radiography series to the case rather than leaving them on the scanner brings storage, viewing and integrity work that is a project in itself.

Photograph volume over a long horizon. A busy office generates an archive that must stay accessible and demonstrably unaltered for decades, and decades of storage and migration is a running cost rather than a build cost.

Dictation. If pathologists will not type, and most will not, transcription and structured capture from dictated narrative is real scope.

Then the recurring number nobody puts in a county budget request. Maintenance runs 10 to 20 percent of build cost per year, so roughly $16,000 to $32,000 on a $160,000 first release, covering hosting, monitoring, archive integrity and the annual drift in laboratory and state formats. Offices that fund a build from a capital allocation and forget the running line end up with a system that ages out in five years.

What separates a build that works from one that fails here?

Deferring the state death registration interface out of release one. It is the single largest scope reduction available, it costs the office very little operationally because the step becomes review and paste rather than re-authoring, and it removes the one dependency you cannot control from your critical path.

Then offline scene capture in the first release rather than as an enhancement. Scene work happens in basements, fields and parking structures, and a system that needs signal gets bypassed with a paper form that somebody types in later, which reinstates the problem you paid to remove.

Then bringing the chief investigator and the administrative lead into design, not just the chief medical examiner. The investigator knows what happens at two in the morning, and that is the part the software either survives or does not.

Then the contract. The repository, the hosting accounts and the case data are yours before kickoff, together with an explicit answer to how the archive would be extracted in bulk if you replaced the system. These records outlive vendors, and often outlive the people who commissioned them.