A county public health department in the upper Midwest has completed a twelve-month cold-chain monitoring pilot using Fieldmesh gateways and environmental sensors across 74 storage sites: clinics, mobile units, two warehouses, and a number of small partner sites with a single fridge in a back room.

The pilot closed in December. Across the year there were 214 temperature excursions, every one of which has a documented cause. That is the result the department was after. Not zero excursions, which would have been implausible, but zero excursions that nobody could explain.

What they had before

Data loggers with a USB port. Someone visited, downloaded a file, and looked at it. At the larger sites that happened weekly. At the small partner sites it happened when someone remembered.

The consequence is not that problems went unnoticed. It is that they were noticed late, and by then the question "was this stock still good on Tuesday?" could not be answered, so the stock was discarded.

The network problem

Twenty-two of the 74 sites do not have reliable connectivity. Some are rural, some are in basements, and one is a mobile unit that is in a different place every day.

Fieldmesh handles this by not requiring the network to be there. Sensors record locally and gateways store and forward, holding up to 90 days of readings and reconciling when a connection returns. A reading taken during an outage arrives late, with its original timestamp, flagged as delayed. It is never lost and it is never silently backdated.

For the mobile unit this turned out to be the whole value of the system. It spends most of its day out of coverage and syncs when it returns to the depot.

Pilot summary

Sites
74, including 22 with unreliable connectivity
Monitored units
310 refrigerators and freezers
Duration
12 months
Excursions recorded
214
Unexplained excursions
0
Longest offline period reconciled
19 days
Stock discarded on a precautionary basis
Down 61 percent year on year

The 3am problem

Six weeks in, one clinic started generating an alert at roughly 3am, two or three nights a week. The fridge would rise about 1.5 °C above its band and return to normal within twenty minutes. Nothing was ever out of range long enough to affect stock, and the pattern was too regular to be a door left open.

It was the defrost cycle on a chest freezer on the same circuit, which had been doing this for years. Nobody had ever seen it because nobody had ever looked at that fridge's temperature at 3am on a Tuesday.

That is the honest summary of what continuous monitoring does. It does not usually find a catastrophe. It finds a handful of small, long-running problems that a weekly download was never going to surface, and it lets you decide which of them are worth fixing.

We expected to catch failures. What we actually got was the ability to say, with a straight face, that stock was in range the whole time. That is the conversation that used to cost us a freezer's worth of vaccine.

Program lead, county public health department

What happens next

The department has moved from pilot to standing deployment and is adding 30 further sites this year, including partner pharmacies. The county has asked for a shared read-only view for those partners, which is a fair request and is work we have scheduled rather than work that exists today.