QC targets (ADM export)

Run this study#

Your question: what QC target mean and SD do your existing QC records support for each instrument, lot, analyte and level?

Bring: the original ADM text exports (.txt or .csv, 1–10 files, up to 2 MB and 50,000 result rows in total) and a baseline period. Each result needs instrument, lot, analyte, date, time, level and result. You can also add a unit for each analyte, and a QC name and expiration date for each QC material. They print on the report.

How much data? Each instrument, lot, analyte and level is a separate group. A suggested starting plan is one QC result per day for 20 days, 20 results per group. Existing records work if they cover that period and the system was stable. Repeats on the same day add results but not days. A group with fewer than 20 results or 20 days is marked with an asterisk and the note Limited baseline. The calculation minimum is two numeric results. The WHO laboratory quality management system handbook, pp. 81–83, describes collecting repeated control measurements to set the mean and SD.

1. Decide the baseline period. Use one stable period for each instrument and QC lot.
2. Upload the original exports. The example column layout shows the seven columns. In the menu, open Tools → QC targets (ADM export). Check the Workup title, choose the files and choose Analyze files.
3. Set the baseline. Under Baseline period, enter the Start date and End date, or choose Use an export’s date range.
4. Review each group. Choose Review. Confirm the Unit and Target precision. Check the time plot and Original observations for shifts, duplicates and flagged results. Give a reason for every exclusion.
5. Save and download. Add the QC name and expiration date under QC materials & report details, choose Save review, then download the Summary CSV and report.

You get: candidate target means, SDs, CVs and optional SD ranges for your laboratory to review and approve. The unit you enter only labels the results. Values are used as exported.

Purpose#

This is a workspace of its own, opened from Tools. It works outside projects, and earlier workups are listed under Saved workups on the same page. Each analyte gets one comparison table across instruments, lots and levels. The group review names the module: Immunoassay for instruments with _IM_ in the name, Chemistry for _CC_.

When to use it#

Use it to set targets for a new QC material or to review an existing QC baseline across instruments, lots and levels.

Study setup#

After Analyze files, the workup opens with every date pooled. The settings below take effect when you choose Save review.

1. Baseline period. Start date and End date are both included. Use an export’s date range fills them from one export, and All dates clears them.
2. Reporting for {analyte}. In a group's review, set Target precision and Unit. Precision is Source decimal places by default, or choose Decimal places or Significant figures. The setting applies to that analyte on every instrument and level.
3. QC materials & report details. Choose Edit details. Enter the QC name and expiration date for each QC family. Lots that differ only in a last digit matching the level form one family, for example 85391, 85392 and 85393 for levels 1 to 3. 85391T to 85393T are a separate family. Tick Include estimated SD ranges in the report to add the ±1, ±2 and ±3 SD table and chart lines.
4. Combined performance. With two or more instruments in the workup, tick two or more that run the same QC. The report then adds a Combined performance table.

Filter the groups with Find an analyte, Instrument, Lot, Level and Review state.

Entering data#

Upload UTF-8 text exports with values separated by comma, tab or semicolon. A header row with these names is optional. The seven columns must come in this order:

Result cells are read as:

• Number: decimal point, plain or scientific, such as 140.0 or 1.20e-3.
• Censored: <, >, <=, >=, ≤ or ≥ followed by a number, such as <3.100.
• Missing: empty, NA, N/A, NULL, NONE, NaN, - or --, any case.
• Unrecognized: anything else, such as hemolyzed or 1,23.

The import stops at a malformed row, a bad date or time, the wrong encoding, the same file twice or a file over the limit, and names the file and line. Uploading the same set of files again reopens the saved workup.

A group with missing, censored or unrecognized results gets a candidate target once each of them is excluded with a reason.

Statistics#

Each group is calculated over the baseline period.

• Target mean: mean of the numeric results, rounded to the target precision. Halves round away from zero.
• Numeric mean: the mean to one more decimal place than the target, printed on the time plot. Saved unrounded values and range in group review keeps the exact saved values. The summary CSV uses up to 17 significant digits for its numeric mean, SD, CV and range columns.
• Sample SD: sqrt(sum((value − mean)²) / (n−1)), from n = 2. Shown to one more decimal place than the target, for example 0.031 for a target of 0.93.
• CV %: 100 × SD / mean, from n = 2 and for a positive mean. Two decimal places.
• Estimated SD ranges: mean ±1, ±2 and ±3 SD from the unrounded mean and SD, shown to one more decimal place than the target.
• Combined performance: for each analyte, lot and level the ticked instruments share, each instrument counts equally. Combined mean = average of the instrument means. SD = √(average of the instrument variances). CV % = 100 × SD / combined mean. Spread % = 100 × (highest − lowest instrument mean) / combined mean. It is calculated only when every instrument in the combination has a candidate target, otherwise the row reads Not combined. A large spread means the instruments read differently, and a shared target would be off-centre for each.

Every result has equal weight, so days with more repeats count more. The SD includes all variation in the period.

Target precision. Source decimal places uses the resolution of the original text. If that varies within a group, the coarsest is used and the group gets an asterisk. The other settings change only the rounding shown.

Candidate target. A group needs at least two numeric results and no unresolved missing, censored, unrecognized, duplicate or conflicting-timestamp rows. Otherwise it shows a dash. An asterisk marks a group that has a note to review.

References: NIST sample SD, NIST coefficient of variation and NIST outlier review.

Worked example#

Upload adm-qc.csv, a synthetic eight-row file, with no baseline dates and source decimal places.

The export's analyte NA appears as Sodium, lot 00123. The immunoassay group has only two <3.100 results at level Low, so n is 0 and it has no target. The asterisks mark the short baseline.

With P_CC_3 and P_CC_4 ticked under Combined performance, Sodium lot 00123 level 1 combines to n = 6, combined mean 140.0, SD 1.58 (√((1.00² + 2.00²) ÷ 2)), CV 1.13 % and spread 0.00 %. Both analyzers have the same mean.

Reading the results#

The comparison table has one row per instrument, lot and level with n, Target mean, SD and CV %. Review opens a group:

• row counts by state, such as included, excluded, missing, duplicate or outside the period
• a time plot of numeric results with the mean and ±1/2/3 SD lines, excluded rows as crosses and repeated timestamps in amber
• Estimated SD ranges and Saved unrounded values and range
• Original observations, one row per source row with its State. Tick Issues and exclusions only to hide clean rows.

A row can be in more than one count, for example both excluded and outside the period, or both duplicate and conflicting timestamp. Identical rows may be genuine repeats, because the export has no observation ID.

To exclude a result, choose Exclude with reason, enter the reason and choose Apply to review. Include again reverses it.

Edits are kept in the page until you choose Save review. Until then, tables, plots and downloads show the last saved review. Discard review edits drops them. If another tab saved first, Load latest & keep my edits merges both.

Saved results & source files offers Summary CSV, Open report, Report PDF and the original files, each with its SHA-256 checksum.

When the report layout has changed since a workup was saved, the save button reads Update report and redraws the report from the same results. When the calculation has changed, it reads Update estimates and recalculates with the current method.

Report. It opens with the baseline, measured dates, instruments, source, result counts and a summary of how many groups have a candidate target and how many need review. Candidate targets lists each instrument, lot and level, grouped by analyte, with lettered notes. A group without a target reads Not assigned. Combined performance and Estimated SD ranges follow when chosen, then QC materials and a Review and approval block with blank lines for reviewer, signature and date. The Levey-Jennings charts, one per group, come next, then the methods and, for several exports, a list of source files.