PrevIQ Docs
Open app

Sample-size tiers & confidence

How PrevIQ tells you whether a prevalence number is solid enough to act on.

A prevalence rate based on 4 specimens and one based on 4,000 specimens look identical on screen — 25% is 25%. PrevIQ never lets them look equally trustworthy. Two mechanisms make precision visible: sample-size tiers and confidence intervals.

Sample-size tiers

Every rate is tagged A, B, or C based on its tested count (the denominator):

Tier Sample size (n) Clinically reportable? Guidance
A n ≥ 100 ✅ Yes Solid. Use for clinical and reimbursement decisions.
B 30 ≤ n < 100 ✅ Yes Reportable, but note the wider confidence interval.
C n < 30 ❌ No Exploration only. Not defensible on its own.

Important

Tier C is never clinically reportable, no matter how high the apparent rate. A 100% rate on 3 specimens is not evidence of anything. PrevIQ surfaces Tier C rows (styled muted/italic) so you can see the full landscape, but they're excluded from clinical claims and always classified as RARE for panel design.

The 95% confidence interval

Alongside each rate, PrevIQ reports a 95% confidence interval — the range the true rate plausibly falls within, for example:

Example

prevalence_rate = 15.2%
95% CI          = [12.3 – 18.4%]

A narrow interval means a precise estimate (usually large n). A wide interval means you should be cautious — the point estimate could be well off.

Why Wilson, not the simple formula

PrevIQ uses the Wilson score interval rather than the textbook normal-approximation (Wald) interval. The Wald interval misbehaves exactly where lab data lives — small samples and rates near 0% or 100% — where it can produce impossible bounds (below 0% or above 100%) and understate uncertainty.

The Wilson interval stays sensibly bounded within [0%, 100%] at any sample size and any rate, which makes it the honest choice for prevalence.

Note

You don't need to compute anything — PrevIQ attaches the Wilson interval to every reportable rate automatically. This explanation is here so you can confidently answer "how was the interval calculated?"

Reading a row in practice

  1. Glance at the tier

    A or B → you can use it. C → explore only.

  2. Glance at the interval width

    Narrow → trust the point estimate. Wide → lead with the range, not the single number.

  3. Then read the rate

    With tier and interval in mind, the rate becomes a number you can defend.