Quality & InspectionFree calculatorPDF, Excel, Word exportMedical Devices

Medical Device Complaint Rate Trending Calculator

Complaint trending is a routine input to management review, CAPA and post-market surveillance. This calculator turns raw complaint counts and units shipped into a rate per 1,000 units and tells you whether the period is above a statistical control limit.

Your numbers

units
21 complaints
per 1,000 units

From the prior 12 months or from your risk file's expected rate.

Your results

Complaint rate per 1,000 units
0.875
Signal (1 = above limit)
0
Expected complaints at baseline
14.4
Upper control limit (count)
25.7842
Poisson approximation: mean plus sigma times the square root of the mean.
Change versus baseline
45.8%

A statistical signal is a trigger for investigation, not proof of a defect. Reportability decisions for any individual complaint follow your regulatory procedures and are independent of trend results.

Save results:

Complaint rate and limit as complaints vary

Sweeps Complaints received in period from 0 to 60, holding your other inputs fixed.

Automate complaint trending

Want complaint counts and shipped units joined from your ERP each month? We can review how.

Optional. The tool and every export work without this. No spam.

How to use it

Enter units shipped and complaints received for the same period, then the baseline rate you expect. The tool returns the rate per 1,000 units, the expected count and an upper control limit. A signal value of 1 means the count is above the limit and the trend should be investigated. Use a 2 sigma limit for early alerts and 3 sigma for action triggers, and document the choice in your trending procedure.

The statistics

Complaint counts are rare events and behave roughly like a Poisson process with mean equal to the baseline rate times units. The standard deviation is the square root of the mean, giving an upper limit of mean plus k times sqrt(mean), the same logic as a c chart. For small expected counts below about five, the normal approximation is weak, so review borderline results manually.

  • Rate = complaints / units x 1,000
  • Expected = baseline x units / 1,000
  • UCL = expected + k x sqrt(expected)

Common mistakes

Using shipments from a different period than the complaints misstates the rate because devices fail with a delay after distribution. Mixing product families with different use environments hides signals. Counting duplicate complaints about the same event inflates the rate, and counting only investigated complaints deflates it. Trend by failure mode and by lot as well as in total, since a lot-specific problem can disappear in a family-level average.

Frequently Asked Questions

What denominator should I use?

Use units distributed in a comparable period, or better, units in service if you can estimate them. Match complaints to the right product family and time window, and state the choice in your trending procedure.

Does a signal mean a CAPA is required?

Not automatically. It means the trend needs review. Your procedure should define how signals are evaluated, for example by checking for clusters, lots and failure modes, and when a CAPA or health hazard evaluation is opened.

Is Poisson suitable for low volumes?

It is a reasonable model for rare independent events, but with expected counts under five the approximation is rough. Use exact Poisson limits in a statistics package or review the trend by judgement alongside the numbers.

How often should I trend?

Monthly or quarterly is common, with additional reviews after launches or design changes. The frequency should be risk-based and stated in your post-market surveillance plan.

If complaints are logged in one system and shipments in your ERP, we can help review how to produce this trend automatically.