Recertification intervals should follow skill-decay rates, not fixed years

What it is

A risk-based recertification interval schedules reassessment from skill-specific retention evidence, operational exposure, failure consequence, and change, rather than assuming every competence expires after the same number of years.

A fixed interval — "refresh every two years" — looks fair and is easy to administer, but it smuggles in two untested assumptions: that every skill a given person holds decays at the same rate, and that a given skill decays at the same rate across different people. The two rarely hold at once.

Why it happens

Skill components genuinely decay at different rates: frequently used actions with simple steps and immediate feedback stay sharp; low-frequency, multi-step actions triggered only by abnormal conditions — certain emergency-response sequences, for instance — decay fast precisely because they are rarely exercised, and their erosion goes unnoticed until they are needed. A fixed cycle re-certifies the frequently used skills redundantly, since those people are already practicing continuously and the recert is a formality; meanwhile the rare, high-consequence skill sits unexercised for the entire interval and is only found to have dropped below a safe threshold when it is actually called on.

Moving from a fixed calendar to a decay-rate basis requires skill-level decay curves first: which specific actions show a measurable drop in pass rate after how much time out of use. The typical source for this is not a dedicated decay experiment but the pre-refresher baseline check — having the trainee run a standard task unassisted before any instruction is given, then correlating that baseline score against how long it has been since the person last performed the task operationally or last refreshed. If baseline data shows one skill's pass rate drops noticeably after six months idle while another holds up after two years, applying the same interval to both means refreshing the first too late and the second too often — resources are not going where the risk actually is.

Studying it

Estimate skill-specific retention curves from longitudinal performance data, real operational exposure frequency (how often the action is actually performed), and historical pre-refresher baseline scores; compare fixed-interval and risk-triggered policies on missed-degradation rate (competence dropping below threshold before recert catches it) and training burden (frequency, time, cost). The baseline check has to run before formal instruction and without advance notice that it will be scored, or it measures a last-minute cram rather than unprompted retained performance. With sparse samples, state uncertainty as a range rather than manufacturing a precise interval to look rigorous.

Where it stops holding

The method depends on accumulated history — newly introduced equipment or a newly rolled-out procedure has no pre-refresher baseline data at all, so there is nothing to fit a decay curve to, and a risk-based interval simply cannot be derived yet. In that situation, fall back to a conservative fixed interval as a floor while systematically collecting baseline data from the first refresher onward, shifting to evidence-based intervals as data accumulates rather than waiting for complete data before designing the process at all.

Regulatory maximum intervals remain a hard constraint that an evidence model can only shorten, never relax on the grounds that "our data says longer is fine." A single passing baseline check does not prove long-term retention — it only shows the threshold was not crossed at that moment, not that it will hold until the next scheduled review. Health changes, organizational restructuring, and equipment modification can all warrant an earlier review regardless of where the person sits in the regular cycle.

Applying it

• Define an observable performance threshold, an exposure record (how often the action is actually performed), and early triggers (health change, role change, equipment modification) for every critical capability.
• Update each capability's interval from rolling baseline performance while keeping a conservative maximum interval as a floor, so a skill that merely "looks stable" in limited data does not drift to an unbounded cycle.
• For rare capabilities with too little real exposure to build a performance history — major emergency response, for example — use periodic simulation instead; an absence of incidents is not evidence of retained competence, since it more likely means the capability was never actually exercised.
• Plot each baseline score against time-since-last-use; when one skill crosses the threshold after a shorter idle period than others, shorten that skill's interval specifically rather than shortening every interval across the board.

Related

• Same group: Y6.04.1 Qualification must match the equipment version actually operated · Y6.04.3 Training records must be traceable for accident investigation · Y6.04.4 Systems should automatically restrict access when qualifications expire
• Nearby: Y6.02 Skill retention · Y6.01 Simulation training
• Search terms: risk-based recertification · performance-based qualification · retention curve