Guide · Operations
Why Hospital Work Breaks Between Departments
Watch a hospital for a day and the pattern is hard to miss. Inside a department, work is competent and practised. Between departments, it depends on somebody remembering.
This is not a criticism of hospital staff. It is an observation about where the software stops.
Departments are competent. The gaps are not owned.
Reception knows how to register a patient. The laboratory knows how to run a sample. The pharmacy knows how to dispense against a prescription. Each of these has been done many thousands of times by people who are good at it, and the software supports each of them reasonably well.
Then the patient leaves one department for another, and the process that was carefully specified inside each becomes an assumption between them.
The order exists in the doctor’s system. Whether the laboratory knows about it depends on whether anybody told them. The result exists in the laboratory’s system. Whether the doctor sees it depends on whether they go looking.
Nobody owns the gap, so it gets covered by whoever notices first: a phone call, a runner, a message in a group chat. It works — and it leaves no record, so when a visit goes wrong nobody can reconstruct where.
The four ways a handoff fails.
Almost every between-departments failure we watched was one of these, and each has a different fix.
The receiving department never learns
Work is created in one system and never surfaces in another. The order exists, correctly recorded, and the department that must act on it has no idea. Discovered when the patient asks why they are still waiting.
The originator is never told it is done
The reverse. A result comes back, sits in a report queue, and the doctor who ordered it is not told. Discovered at the next appointment, or not at all.
Context is dropped in transit
The patient arrives, but what was already known does not. Vitals taken forty minutes ago are re-taken. History given at reception is asked again. Each repetition costs time and creates a chance for the two records to disagree.
The patient is lost between queues
They leave one queue and, for a while, are in none. Time already waited is not carried. If they step away, they are either skipped or hold up a room while somebody looks for them.
The phone call is a symptom, not a workaround.
Every hospital has developed a human protocol to cover these gaps, and it usually works. But it scales with staff attention rather than with volume, it degrades on the busiest days — exactly when it matters — and it disappears at shift change.
It also leaves nothing behind. A hospital that coordinates by phone cannot answer, afterwards, when the laboratory was told, or why a result waited two hours.
When a doctor finishes a consultation, the next team should not need a phone call to know there is work waiting.
What software should carry instead.
The fix is not more notifications. It is making the handoff a thing the system owns rather than an assumption between two systems.
- Completing work should create the next work. An order placed during a consultation should appear on the receiving department’s worklist as a consequence of finishing the consultation, not as a separate act of telling somebody.
- A result should return to the person who asked for it, as work to review, rather than into a queue that must be visited.
- Context should travel with the patient, because it is attached to the visit rather than to the department that recorded it.
- A patient in transit should still hold a place. Leaving one queue for another should not reset what they have already waited.
- Every handoff should leave a timestamped record, so the question "when did the laboratory know?" has an answer rather than a recollection.
How to test this in your own hospital.
You do not need software to run this diagnostic, and it takes an afternoon. It is also the most useful preparation for evaluating any vendor.
Follow one patient
Walk with a single outpatient from the front door to checkout. Note every point where information is carried by a person rather than by a system.
Count the re-identifications
How many times is the patient asked who they are, or their details typed again? Each one is a place two records can diverge.
Time one order
From a doctor placing a laboratory order to the laboratory acting on it. Then ask how they found out.
Ask about yesterday
Pick a visit from yesterday and try to reconstruct it: when was the sample taken, when was the result available, when did the doctor see it? Note how much comes from memory.