Delta Check Calculator
Delta Check Calculator
The change between a patient’s current result and their previous one, absolute and proportional, against the time between them — and against limits only your own laboratory can set.
Delta Check
Two results + interval → deltaA creatinine of 128 µmol/L today against 80 µmol/L two days ago, in a laboratory using Whitehurst’s creatinine rules — a 25% percentage limit and a 44 µmol/L absolute limit
The four delta calculations
percentage delta = |current − previous| ⁄ previous × 100%
absolute rate delta = absolute delta ⁄ interval in days
percentage rate delta = percentage delta ⁄ interval in days
an alert fires when the chosen figure exceeds a limit your laboratory has set for this analyte
- which of the four
- all four are in published use. Absolute deltas suit analytes with a narrow range and a meaningful unit change, such as sodium; percentage deltas suit analytes that span decades, such as creatinine or an enzyme; the rate forms add the interval and are the only ones that distinguish a rise over six hours from the same rise over six months
- there is no universal limit
- this is the page, not a caveat on it. Delta check limits are analyte-specific and laboratory-specific, because they depend on the analyte’s within-subject biological variation, on your own imprecision, and on the population you serve. A laboratory serving a dialysis unit and one serving a general practice cannot use the same creatinine delta limit without one of them drowning in alerts
- how limits are actually set
- three approaches are in use. Take a percentile — the 95th or 99.5th — of the distribution of deltas your own laboratory already sees, which is empirical and self-calibrating. Compute a reference change value from biological and analytical variation, which is principled and usually produces limits that alert too often. Or choose from clinical experience and adjust after a period of use, which is what most laboratories end up doing
- what it is for
- the two purposes stated in the literature are to detect specimen and collection problems that quality control cannot see — a mislabelled tube, a drip-arm sample, a mix-up at the bench — and to draw attention to a clinically significant change. The first is the reason delta checks exist; the second is what they mostly deliver
- the predictive value problem
- specimen misidentification happens in roughly 0.04% to 1% of samples, and later studies put it below 0.1%. A rule with excellent sensitivity applied to an event that rare produces mostly false alarms, because real physiology trips the same rule far more often. Published positive predictive values for single-analyte delta checks run from 0.44% for urea and creatinine through 3.3% in one Japanese series to about 12% at a realistic error rate — meaning between eight and two hundred alerts investigated for each mix-up found
- the reference change value
- the related but different idea. The RCV asks how much a result must move before the change is unlikely to be noise, from analytical and within-subject biological variation: √2 × Z × √(CVa² + CVi²). A delta check asks whether a change is large enough to be worth a human looking at the specimen. An RCV makes a defensible starting point for a delta limit and usually needs widening, because it was designed to detect real change and will therefore flag a great deal of it
- check both specimens
- a delta check compares today’s result with a previous one, so it fires when either is wrong. If the previous sample was the mislabelled one, today’s correct result triggers the alert and the investigation starts in the wrong place
Worked example
A creatinine of 128 µmol/L today against 80 µmol/L two days ago, in a laboratory using Whitehurst's creatinine rules — a 25% percentage limit and a 44 µmol/L absolute limit
Absolute change = |128 − 80| = 48 µmol/L, a rise
Percentage change = 48 ⁄ 80 × 100 = 60.0%
As rates over the two days: 24 µmol/L per day, or 30% per day
Against the limits: 60.0% exceeds the 25% percentage limit, and 48 µmol/L exceeds the 44 µmol/L absolute limit. Both rules fire
Now the part that matters. The alert is correct and the commonest explanation is not a specimen error at all: a 48 µmol/L rise in creatinine over 48 hours is stage 2 acute kidney injury by the usual criteria, and acute kidney injury is far commoner than specimen misidentification
So work in order. Look at the rest of today's profile for internal consistency — a mislabelled tube usually disturbs several analytes at once in unrelated directions, while acute kidney injury moves urea and creatinine together and leaves the full blood count where it was. Then check the patient's clinical course. Then, if neither explains it, check the identity of both specimens
And note what the numbers cannot tell you. Specimen misidentification occurs in well under 1% of samples, so even a rule with good sensitivity generates mostly false alarms: published positive predictive values for single-analyte delta checks run from 0.44% to about 12%
Published delta check limits, each with its source
| Analyte | Limit | Interval | Source |
|---|---|---|---|
| Sodium | ±3.1% (reference change value at 95%) | — | Lee et al 2016, the smallest of nine chemistry tests over 1.5 million paired results |
| Sodium | 5% (percentage delta) | — | Ladenson’s scheme, reported by Randell and Yenice 2019 |
| Sodium | 13 mmol/L | 3 days | Worked example in the ADLM/AACC Scientific Short on delta check limits |
| Creatinine | 44 µmol/L absolute when the current result is below 177 µmol/L, else 25% | — | Whitehurst, reported by Randell and Yenice 2019 |
| Creatinine | Rise > 26 µmol/L | 30 days | Garner et al, reported by Randell and Yenice — an acute kidney injury rule, not a mix-up rule |
| MCV | 3 fL | 3 days | Strathmann et al 2011 — the best-performing single analyte in that simulation |
| Troponin T | 50% | 3 hours | Reported by Randell and Yenice 2019 |
| ALT | ±67.6% (reference change value at 95%) | — | Lee et al 2016, the largest of the same nine tests |
| Most tests | around 20% | — | Ladenson’s scheme, reported by Randell and Yenice 2019 |
How well delta checks actually detect a specimen mix-up
| Study | Finding | What it means |
|---|---|---|
| Schifman et al, CAP Q-Probes | “only 0.3% of all delta check alerts were related to specimen misidentification” | About 300 alerts investigated for every mix-up found |
| Strathmann et al 2011 | Positive predictive value 0.44% for urea and creatinine, 4.1% for MCV | MCV was the best single analyte and still gave 24 false alarms per true one |
| Iizuka et al | “only 3.3% of delta check alerts represented true positives” | Detection was good; the alerts were mostly something else |
| Furutani et al | PPV 58.2% at an assumed high error rate, 12.1% “at a more realistic error rate of 0.1%” | The predictive value is driven by how rare mix-ups are, not by how good the rule is |
| Base rate | Specimen misidentification in 0.04% to 1% of samples, under 0.1% in later studies | This is the number that makes the arithmetic unforgiving |
What a delta check is for, and why nobody can tell you the limit
A delta check compares a patient’s current result with their previous one and raises an alert when the change is larger than some limit. It exists for a reason quality control cannot cover: a control chart in perfect order says nothing about whether the tube in the analyser belongs to the patient whose name is on it. Mislabelling, a mix-up at the bench, a sample drawn from above a drip — none of these disturb the controls, and all of them produce a result that is wrong for the patient it is reported against. The literature gives delta checks two purposes: to identify specimen and collection problems that quality control methods do not identify, and to assist clinicians in recognising significant change in a patient’s status. The first is why they were invented, and the second is what they mostly deliver.
There are four ways to compute the change and all four are in published use: the absolute difference, the percentage difference, and each of those divided by the interval between the samples. Which to use is an analyte question. An absolute limit suits something with a narrow physiological range and a meaningful unit step, such as sodium; a percentage suits something that ranges over decades, such as creatinine or a transaminase; and the rate forms exist because a 60% rise over six hours and the same rise over six months are not the same event. Many laboratories run two tiers, an absolute limit at low concentrations and a percentage at high ones, because a percentage of a very small number is a very small number.
What no one can give you is the limit itself, and this is not a hedge. Delta check limits are analyte-specific and laboratory-specific because they depend on the analyte’s within-subject biological variation, on your own analytical imprecision, and on the population your laboratory serves. Look at what the literature actually contains for a single analyte. Sodium has been given a reference-change-value limit of ±3.1%, a percentage delta limit of 5%, and an absolute limit of 13 mmol/L which on a sodium of 140 is over 9% — three published figures differing by a factor of three, because they were derived by different methods for different purposes. A laboratory serving a dialysis unit and one serving general practice cannot share a creatinine limit without one of them being flooded. The three routes actually used are to take a high percentile of the deltas your own laboratory already sees, to compute a reference change value from biological and analytical variation, or to choose from clinical experience and adjust after a period of use — and most laboratories end up doing the third whichever they started with.
The uncomfortable arithmetic is the predictive value. Specimen misidentification happens in something between 0.04% and 1% of samples, and the later estimates sit below 0.1%. Apply even an excellent rule to an event that rare, in a population where real physiology trips the same rule constantly, and most alerts will be false. The published figures bear that out: a CAP Q-Probes study found that only 0.3% of delta check alerts related to specimen misidentification; a simulation study found positive predictive values of 0.44% for urea and creatinine and 4.1% for MCV, the best-performing single analyte; a Japanese series found 3.3% of alerts were true positives despite good detection; and a further study reported 12.1% at a realistic error rate. Strathmann and colleagues drew the conclusion plainly: the low yield of delta checks based on any single analyte should prompt careful evaluation of their practical utility.
None of which is an argument for switching them off. It is an argument for expecting the alert to be real clinical change, and for working in that order. Check whether the rest of today’s profile is internally consistent — a mislabelled tube usually disturbs several unrelated analytes at once, while acute kidney injury moves urea and creatinine together and leaves the full blood count alone. Check what happened to the patient between the two samples: a transfusion, a dialysis session, fluid resuscitation, a drug started. Only then question the specimen, and question both specimens, because a delta check fires when either result is wrong and if the previous sample was the mislabelled one the investigation has started in the wrong place.
One further distinction is worth keeping clear. The reference change value answers a related but different question: how much must a result move before the change is unlikely to be analytical and biological noise? It is built from the analytical CV and the within-subject biological CV and it is about the patient’s physiology. A delta check limit is an operational threshold for a human to look at a specimen. An RCV makes a principled starting point for one and almost always needs widening, because it was designed to detect real change and will faithfully flag a great deal of it. The gap between the two is where laboratory judgement lives, and it is why the last honest thing to say about delta check limits is that one careful study of haemoglobin found a criterion of 1.0 for outpatients and 4.9 for inpatients — same analyte, same laboratory, same metric, a fivefold difference, decided entirely by who the patients were.
Frequently asked questions
How do you calculate a delta check?
Four ways, all in published use. The absolute delta is the difference between the current and previous results. The percentage delta is that difference divided by the previous result, times 100. Each has a rate form, divided by the interval between the samples in days. A creatinine of 128 µmol/L against 80 µmol/L two days ago gives an absolute delta of 48 µmol/L, a percentage delta of 60%, and rates of 24 µmol/L or 30% per day.
What are the standard delta check limits?
There are none, and that is the substance of the matter rather than a caveat. Limits depend on the analyte’s within-subject biological variation, on your own imprecision, and on the population you serve. Sodium alone has published limits of ±3.1%, 5% and 13 mmol/L from three different sources, all derived legitimately for different purposes. Use published figures for the order of magnitude and derive your own limits from your own delta distributions.
What is a delta check actually for?
Primarily for detecting specimen and collection problems that quality control cannot see — a mislabelled tube, a mix-up at the bench, a sample drawn above a drip — because a perfect control chart says nothing about whether the specimen belongs to the patient named on it. Its secondary use is to draw attention to clinically significant change, and in practice that is what most alerts turn out to be.
Why do delta checks generate so many false alarms?
Because specimen misidentification is rare — between 0.04% and 1% of samples, and under 0.1% in later studies — while real physiological change is common and trips the same rule. Published positive predictive values for single-analyte delta checks run from 0.44% to around 12%, so between eight and two hundred alerts are investigated for each mix-up found. The problem is the base rate, not the rule.
What is the difference between a delta check and a reference change value?
The reference change value asks how much a result must move before the change is unlikely to be analytical and biological noise, and is computed as √2 × Z × √(CVa² + CVi²) from imprecision and within-subject biological variation. A delta check limit is an operational threshold for flagging a result for human review. An RCV is a principled starting point for a delta limit and usually needs widening, because it was designed to detect real change and will flag a great deal of it.
A delta check has fired. What should I do first?
Look at the rest of today’s profile for internal consistency, because a mislabelled specimen usually disturbs several unrelated analytes at once while a real physiological change does not. Then look at what happened to the patient between the two samples — transfusion, dialysis, fluid resuscitation, a drug started. Only then question specimen identity, and question both specimens rather than only today’s: the check fires when either result is wrong.
Related calculators
References
- Randell EW, Yenice S. Delta checks in the clinical laboratory. Crit Rev Clin Lab Sci. 2019;56(2):75–97.
- Strathmann FG, Baird GS, Hoffman NG. Simulations of delta check rule performance to detect specimen mislabeling using historical laboratory data. Clin Chim Acta. 2011;412(21–22):1973–1977.
- Lee J, Kim S-Y, Kwon HJ, Lee HK, Kim Y, Kim Y. Usefulness of biological variation in the establishment of delta check limits. Clin Chim Acta. 2016;463:18–21.
- Kim MS, Park CJ, Namgoong S, et al. Effective and practical complete blood count delta check method and criteria for the quality control of automated haematology analysers. Ann Lab Med. 2023;43(5):418–424.
- Schifman RB, Talbert M, Souers RJ. Delta check practices and outcomes: a Q-Probes study. Arch Pathol Lab Med. 2017;141(6):813–818 — quoted here as reported by Randell and Yenice.
Medical Disclaimer: The tools and content provided here are for educational and reference purposes only. They are not intended to substitute for professional medical advice, diagnosis, or treatment. Clinical decisions should always be based on the comprehensive assessment of a qualified healthcare professional.