Today
A faulty lot of infusion sets first shows up as scattered alarms in many hospitals: an extra air-in-line alarm here, an occlusion without cause there. Nurses insert a new set, and hardly anyone files a complaint. The pumps store every alarm, but the logs are read out only during maintenance or after a complaint and never linked to set lots, because the pump does not know the lot of the set in it. So the signal is noticed only when complaints pile up.
An example
Valbrecht reads out the logs of around 600 VZ 248 pumps during maintenance every month. The agent sees air-in-line alarms in nine hospitals rise from 0.8 to 2.9 per 100 infusion hours, each time in the weeks after deliveries of set lot 2607-B. So far there are only two complaints. The production record shows that 2607-B and two other lots share the same lot of silicone tubing for the pump segment, and these show slightly higher rates too. Quality management puts the remaining stock of the three lots on hold and requests retained samples; measurements show a wall thickness at the lower tolerance limit. The responsible person uses the rates to assess whether a trend report is needed, and the supplier receives a complaint with all the data.
Your data, wherever it lives
The clues to a bad lot are widely scattered today: alarm logs in the pumps and as read-out files in the service system, deliveries and lots in the ERP, complaints as tickets and e-mails, production records and test reports in manufacturing. demi reads, understands and links them per pump, hospital and lot, with your rights. AI spots a conspicuous lot within weeks instead of months; your quality management and the responsible person decide on holds and reports.
Where the data lives today
- Machines & sensors
- QM system
- ERP
- Tickets
- Databases
- E-mails
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: device logs, the service system, the ERP, complaint handling and production records. demi replaces nothing and links every alarm via the serial number to pump, hospital and ward, and via delivery notes to the set lots in use there at the time. Complaints, returns, pump software versions and each lot's production record down to the raw material lot come on top.
- 2 · PermitPermission Layer
Device data and hospital names are visible to service, quality management and regulatory affairs; sales sees only released analyses. Patient data stays out: the analysis covers alarms, run times and settings, not names. Holds, field safety notices and reports to the authority go out only after release by the person responsible for regulatory compliance, and every decision is logged.
- 3 · ActIntelligence Layer
The agent brings logs from every software version into one format, computes alarms per 100 infusion hours and assigns them, via delivery date and the ward's consumption, to the lots in use at the time. With statistical tests it compares every lot with the usual level and reports a deviation with hospitals, alarm type and period. It adds complaints and the production record and shows which lots share the same raw material lot.
How it runs
- 1
Pump logs read out during service or by remote access come in, whatever the software version.
- 2
demi assigns every alarm via serial number and deliveries to the hospital, the ward and the set lots in use there.
- 3
The agent computes alarm rates per lot and reports a statistically conspicuous lot before complaints pile up.
- 4
For each finding it compiles complaints, returns, the production record and lots sharing the same raw material lot.
- 5
Quality management and the responsible person assess it, place holds where needed and decide on reporting and field safety notices.
What changes
A conspicuous lot noticed within weeks instead of months
MDR trend reports based on solid rates rather than single cases
Complaints, logs and production records in one finding
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+144 hfreed per year
Assumptions and calculation
demi maps alarms from device logs to set lots via deliveries; you assess the finding.
60 min → 42 min per case
How we calculate
40 complaints and log analyses × 60 min × 30% × 1.0 (manual) × 12 months = 8,640 min = 144 h
Value per year, before demi costs
Freed per year+144 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€9,360
gained per year
144 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
How does demi get the device logs?
Through the routes that already exist: logs that service reads out during maintenance and safety inspections, logs from complaints and, where hospitals allow it, data from remote service. demi reads the formats of every software version and assigns each entry to the right pump via its serial number.
The pump does not know the set lot. How does demi assign it?
Through deliveries: the ERP shows which lot went to which hospital when, and consumption shows which lots were in use there at what time. The agent works with these time windows and shows how certain an assignment is. Where a hospital names the lot in its complaint, the assignment becomes exact.
Is patient data involved?
No. The analysis covers alarms, run times, flow rates and software versions, not names or health data. If a log does contain personal entries, demi masks them on import, and access to hospital data is logged.
Does this replace vigilance and trend reporting?
No, it gives them numbers. The MDR requires statistically significant increases to be reported, and these are exactly the rates demi computes per lot, alarm type and period. Whether an incident or a trend must be reported is decided by the responsible person; the agent prepares the report with all evidence.
