Two Analysts, Two Fluoride Maps: Which List Goes to the Health Office
A regional water-quality laboratory is mapping fluoride in 40 shallow groundwater wells in a dry-zone district, using one ion-selective electrode and one meter. Two analysts ran the same well list a week apart. Analyst A reports a mean of 1.18 mg/L (range 0.32–2.40); Analyst B reports 0.71 mg/L (range 0.10–1.61). Both used the same fluoride standards and the same calibration slope. On inspection: A buffered every sample with TISAB and read at pH 5.5–6.0; B buffered only the standards and read samples that had been acid-preserved in the field (pH about 2) without further treatment. Paired wells differ between the two runs by 40–70%, and the discrepancy is largest in samples with high iron and aluminium. The national guideline is 1.0 mg/L. The district health office has asked for the list of wells 'safe for drinking' so that tube-well construction can be prioritised this quarter. B's list would clear 31 of the 40 wells; A's would clear 19.
1. State which data set is more likely to be biased and in which direction, justify it from the chemistry, and design a confirmation experiment that would settle the question within one week.
Read the worked answer
Which run is biased, and which way. B's, and it reads low. The electrode responds to free fluoride ion activity, not to total fluoride. B's samples were acid-preserved (pH about 2) and read unbuffered, so a large fraction of the fluoride is protonated as HF and much of the remainder is complexed by Al³⁺, Fe³⁺ and silicate. The electrode sees only the small residual free F⁻, so the reading falls, and the deficit grows with iron and aluminium - exactly the pattern observed. The identical standards and slope rule out a calibration-side explanation, so the cause must be on the sample side.
Why the direction matters. A low bias manufactures false 'safe' classifications. For a health-protection decision that is the more dangerous of the two possible errors.
One-week confirmation design
- Select 12 wells stratified by iron: four low (<0.3 mg/L), four medium, four high (>3 mg/L), deliberately including the wells where A and B disagree most.
- From one unpreserved subsample per well, prepare three aliquots: (i) TISAB added immediately, pH 5.5; (ii) acidified to pH 2, held 24 h, then TISAB only; (iii) acidified as in (ii), then neutralised to pH 5.5 into TISAB containing CDTA.
- Run a certified reference water, a method blank, and a matrix spike on the four high-iron wells.
- Analyse paired differences (i) − (ii) and (i) − (iii). If (ii) reproduces the 40–70% gap and (iii) closes it, pH and complexation are confirmed. If (iii) still reads low, quantify the residual matrix effect by standard addition.
2. The health office wants its priority list this week. Set out exactly what you release, in what wording, and what you refuse to release.
Read the worked answer
Release something now; withholding everything is also indefensible when two wells may be several times a health guideline. But releasing B's list as a clearance is worse than releasing nothing, because a false assurance cannot be recalled.
What goes out this week
- A's results, explicitly labelled provisional, single determinations, with method, date and the unresolved inter-analyst discrepancy stated in one plain sentence.
- Three bands instead of two: below 0.8 mg/L (no immediate action, confirm next round); 0.8–1.2 mg/L (unresolved, resample within two weeks); above 1.2 mg/L (interim advice to avoid for drinking and cooking while alternative supply is arranged).
- A named contact and a firm date for the confirmed list.
What does not go out
- Any wording that a well is 'safe'. A single value below a guideline is not a safety certificate; write 'below the guideline on one occasion'.
- B's list in any form, including informally to the construction programme.
Process. Put the caveats in a memo signed by you and the QA officer, copied to the Director. If a construction priority list must be issued before confirmation, rank wells on A's values and explicitly defer the high-iron wells - a deferral is honest, a wrong clearance is not.
3. The confirmation run shows acid-preserved aliquots reading 0.45 mg/L lower than buffered aliquots, consistently across 12 wells, with 96% spike recovery in the buffered method; high-iron wells need CDTA-TISAB plus standard addition to reach 97–102% recovery. What is the corrected status of the 31 wells B cleared, and how would you handle a B-type list that had already been published?
Read the worked answer
Recalculate, do not average. The corrected value for each well is the buffered result; the acid-preserved result is a known-biased measurement and must be excluded, not blended in as a replicate. On A's values about 19 wells sit below 1.0 mg/L, so roughly a dozen of B's 31 clearances are withdrawn. Any well whose corrected value lies within about 20% of the guideline (0.8–1.2 mg/L) stays 'unresolved' rather than 'safe'.
If the B-type list is already in circulation
- Issue a formal erratum to the same distribution list, with the same prominence and speed, stating which wells are reclassified and why.
- Ask the health office to notify users of the affected wells directly; a list on a noticeboard is not notification.
- Record the correction in the laboratory's quality system, including the date the error was identified and who was informed.
Root cause and prevention. The gap was a procedure failure, not an analyst failure: the SOP said TISAB for standards and was silent on preserved samples. Revise it - field-preserved samples must be neutralised into buffer before any ISE reading, no exceptions. Add a mandatory matrix spike for every batch containing high-iron samples, a CRM control chart, and a second-analyst check on any survey that changes a regulatory classification.