Metrics That Mean Something

Part of Clinical Project Management (PM) Training: Site Oversight Without Doing Their Job.

About 1 minutes. Free, in your language, no account needed.

Most site metrics measure effort

Visits performed, pages monitored, queries raised. These describe activity and are easy to collect. None of them tells you whether the data is trustworthy or the participants are safe.

Metrics that carry signal

Time from visit to data entry. Rising lag is the earliest visible sign of a site under strain, and it usually appears before any data quality problem does.

Query rate per form, and repeat queries on the same field. A repeated query means the site does not understand what is being asked, which is a training issue rather than a carelessness issue.

Screen failure rate against the protocol's assumption. Far above suggests eligibility is being read differently at that site. Far below deserves a look too.

Protocol deviations by category. Ten deviations of the same type at one site is a process problem. Ten different types is a comprehension problem.

Time from awareness to reporting for safety events. The one metric where a trend is never acceptable to watch passively.

Averages hide the thing you need

A trial-level average of any of these tells you almost nothing, because the distribution is what matters. One site with a thirty-day entry lag and eleven at two days averages to something reassuring and is not reassuring at all.

Always look at the distribution and the outliers. The outlier is the finding.

A metric with no threshold is a number

Before you collect a metric, write down what value would make you act and what you would do. A dashboard nobody has agreed thresholds for produces monthly discussion and no decisions.