Acceptance limits

An acceptance limit is a limit the laboratory sets for one estimate. To add one, see Set acceptance limits. Each study's own rules are under Acceptance limits on its reference page.

Reference interval establishment and LoQ verification have no guided limit. LoQ verification uses the quantitation goal in the detection settings. More statistics in the limit dialog offers every other statistic a study calculates, such as slope and intercept in method comparison, between-day SD in multi-day precision, or minimum and maximum in a descriptive summary. A limit for a statistic the study doesn't calculate, such as one copied from another study type, is flagged on Set up; its outcome is Not applicable.

These studies keep their limits with the rest of their settings on Set up:

A descriptive summary, a new reference interval, reference verification counts, a detection verification share, linearity estimates and stability's tested duration can be calculated without a limit: their Acceptance limits section reads Optional. A study that needs a limit for its decision reads None yet until a limit is added. Method, instrument and lot comparisons need a limit for individual specimens: without one, Set up shows Add a limit for individual specimens before deciding agreement. and Specimen agreement is Undecided, even when mean-bias or regression limits exist.

Bias, a specimen difference, a deviation from the fitted line and an interference effect take a limit in the study unit (for example Limit (± mg/dL)), a percentage (Limit (± %)), or both. A repeatability or within-laboratory SD limit works the same way, with Maximum SD (mg/dL) and Maximum CV (%).

With both parts filled, With both limits decides: Pass when within either (the larger allowance) or Pass only when within both. For specimen differences, Pass only when within both keeps a known failure of the study-unit limit when the percentage cannot be evaluated, such as at a zero comparator result; a difference within the study-unit limit stays Undecided until both parts can be evaluated.

These types are offered under Limit type in More statistics:

For Individual specimen agreement, the types are Absolute difference limits, Percentage of each comparator result (in lot comparison, Percentage of each previous-lot result), Percentage or absolute, whichever is greater and Concentration-dependent limits; the greater allowance is taken at each specimen's comparator or previous-lot result.

For glucose at an assigned value of 50 mg/dL, 8% is 4 mg/dL. The greater of 4 and 6 mg/dL is 6 mg/dL, giving a mean acceptance range of 44–56 mg/dL. At 100 mg/dL, the same rule allows 8 mg/dL, giving 92–108 mg/dL. A decision-level bias check is centred on zero bias: ±6 or ±8 mg/dL at those concentrations.

CLIA / CAP limits apply the larger allowance where the table gives an absolute and a percentage part, as the tables state (whichever is greater). Absolute and combined table limits need the listed unit, in any letter case; high-sensitivity CRP, salicylates and testosterone also accept an alternative unit. Table rows based on a peer-group SD, qualitative rules and dilution rules are not offered as limits.

Offered for the Mean or Median in descriptive summary, repeatability and bias studies, the Mean in multi-day precision, Bias at a medical decision level in method comparison, and Individual specimen agreement in method, instrument and lot comparison. A limit of ±2 mg/dL up to 100 mg/dL and ±2% above reads:

• A range includes its start and excludes its end; the final range includes both. A blank To has no maximum.
• The percentage is taken of the assigned value, the decision level or, for a specimen rule, each specimen's comparator or previous-lot result.
• Ranges are listed in order, each starting where the one before ends. A gap or an overlap between two ranges is refused when the limit is added; a gap's message names the concentrations it leaves uncovered.
• The range containing the assigned value or decision level sets the limit; if none contains it, the limit is Not evaluated. In a specimen rule, each specimen's comparator or previous-lot result picks its range, and a specimen outside every range has no usable limit.

• Applies to is All levels or one level; before any results exist it is a text box, where * means all levels and a level label must match the data exactly. Instrument comparison offers analyzers and stability offers storage conditions. Method and lot comparison limits cover the whole study.
• A ± limit checks both directions. A one-sided study-unit limit is Only too high: at most + the limit or Only too low: at least − the limit. A percentage limit applies in both directions.
• Specimen limits in method, instrument and lot comparison are checked at each specimen's concentration: its comparative or reference-analyzer result, or its previous-lot result in a lot study.
• Bias at a medical decision level takes one limit per decision level. A decision level outside the measured range stays Undecided.
• Limits are typed with a decimal point (1.5), even when the data use a decimal comma. A ± limit is a positive number.

The study status reads Criteria met, Criteria not met, Undecided or Not evaluated; the study menu shortens these to Met, Not met, Undecided and Not evaluated. Report page 1 uses the same words, or reads No applicable criteria or Descriptive results. In the project's study list, a study that needs a limit but has none shows No acceptance criteria set.

The reason sits under the outcome on screen and in a note in the report.

Analysis problems gives the fix for each.

• Decisions use the full-precision statistic, before display rounding. Results are shown to Display decimal places; limits are shown as entered. A value whose rounded display reads as equal to the limit, or on the other side of it, is marked †. The report states its rounding rule in the appendix section Decision rules and display precision.
• Units must match exactly; a unit mismatch gives Undecided.
• A limit for one level is Not applicable to the other levels. A limit whose level matches no level in the data is Not evaluated when another limit applies.
• A limit that lacks the required results, independent specimens or deciding statistic is Undecided. A specimen with no result or no usable limit leaves a specimen rule Undecided, unless the failures already decide it.
• Specimen rules compare each difference with its allowance in exact decimal arithmetic.

The study status combines every limit in every group:

1. Not applicable limits are set aside.
2. If nothing is left, the status is Not evaluated.
3. If no results were included, the status is Undecided.
4. If any limit is not met, the status is Criteria not met.
5. Otherwise, any Undecided or Not evaluated limit makes the study Undecided.
6. The status is Criteria met when every remaining limit is met.

Method, instrument and lot comparison always include a Specimen agreement check. On the five results in descriptive.csv (mean 6.0 mg/dL, SD 3.16 mg/dL, all at level 1):

The observed value is the count, for example 19/20 (95%). A specimen missing a result or a usable limit counts as not evaluated. The rule is met only when every specimen was evaluated, and not met as soon as the failures alone exceed the number allowed: with 20 specimens, 1 failure plus 1 not evaluated is Undecided, and 2 failures plus 1 not evaluated is Not met.

• Qualitative agreement can compare overall, positive or negative agreement at the lower end of its Wilson CI.
• An interference effect is met when its whole CI lies within the limits. A point estimate inside the limits with a wider or missing interval is Undecided.

A result exactly on a limit meets it unless Equality at the limit is set to Not met (< or >).

With a mean of exactly 6 mg/dL, "Mean at most 6 mg/dL" is met and "Mean below 6 mg/dL" is not. In 21-position carryover, carryover equal to the error limit is not met.