Diagnostics as a Service and sample quality: why blood samples fail before testing and which results that changes

When red cells break open inside the tube, a Diagnostics as a Service product reports a potassium result produced by that damage rather than by the customer's blood. The name for that damage is haemolysis. It belongs to a wider subject called preanalytics, which covers everything between the draw and the moment testing starts.

Most preanalytics material we have read claims that the majority of laboratory errors happen before a sample reaches an analyser. The direction of that claim is right. The top of the number usually quoted, though, is above the highest figure the published studies report. This article gives the real figures. It then covers what a damaged sample does to named results, what a finger prick can and cannot replace, and what the collection standard ISO 20658 requires, and of whom.

Where do blood tests go wrong?

Before the analyser, in most of the published counts. Two of the studies quoted most often come from one hospital department in Padova. The first, published in 1997, found 189 errors in 40,490 analyses and put 68.2 percent of them before the analysis.¹

The same group repeated the work ten years later on 51,746 analyses and reported 61.9 percent.² So the second figure is about six percentage points lower than the first, from the same department and the same kind of count.

Across the whole literature the published figures are a range rather than a single number. A review by one of the same authors gives that range as 46 to 68.2 percent before analysis, and 18.5 to 47 percent after it.³

The figure of 60 to 70 percent that gets quoted has a different origin. It comes from a 2011 review whose authors state that preanalytical errors account for nearly 60 to 70 percent of all problems.⁴

That review reports the figure rather than measuring anything. As we read it, the figure is a rounding of the published range, and the top of that range, 68.2 percent, comes from the single study of 189 errors. The range is the more honest figure to quote, and quoting it loses none of the argument.

The two studies behind the claim come from one hospital department in Padova. The 1997 paper found 189 errors in 40,490 analyses, 68.2 percent of them before analysis. The same group's repeat in 2007 found 61.9 percent. The published range across the literature is 46 to 68.2 percent. Sources: Plebani and Carraro, Clinical Chemistry 1997; Carraro and Plebani, Clinical Chemistry 2007; Plebani, Clinical Chemistry and Laboratory Medicine 2006; Lippi and colleagues, Clinical Chemistry and Laboratory Medicine 2011.

How often does a sample fail?

Often enough that it has to be planned for. A Spanish national quality scheme publishes eighteen indicators across its member laboratories. The median rejection rate across all requests in that scheme was 2.18 percent.⁵

Within those rejections, haemolysis is the leading cause. One review puts it at up to 40 to 70 percent of all unsuitable specimens, and at up to 3.3 percent of all routine samples.⁶

Which results does a damaged sample change?

The ones that were inside the red cells to begin with. A study of 48 analytes, with a reference group of about fifteen thousand samples per marker, reported how far three of those results differ as haemolysis worsens. The three are potassium, lactate dehydrogenase and aspartate aminotransferase.⁷

Potassium was 0.24 millimoles per litre higher at the mildest level of haemolysis and 1.21 millimoles per litre higher at the worst. Lactate dehydrogenase was 91 units per litre higher at the mildest level and 442 higher at the worst. Aspartate aminotransferase was 6.5 units higher at the mildest and 27.5 higher at the worst.⁷

Those are adjusted differences drawn from one laboratory's routine data rather than from a controlled experiment.

Three common results as haemolysis worsens, against a reference group of about fifteen thousand samples for each marker. Potassium rises by 0.24 millimoles per litre at the mildest level and by 1.21 at the worst. Lactate dehydrogenase rises by 91 units per litre and by 442. Aspartate aminotransferase rises by 6.5 units and by 27.5. Source: Ma and colleagues, PLoS ONE 2026.

The damage does not have to be visible to matter. An earlier study found clinically meaningful changes in aspartate aminotransferase, lactate dehydrogenase, potassium, sodium and chloride in samples where haemolysis was mild or not visible at all.⁸

The obvious answer to a known shift is a correction factor, but the same authors ruled one out. The response to haemolysis was too varied to correct for statistically, so a correction factor is not something that study supports.⁸ We could not find a later study that establishes one.

Is a finger prick as good as a vein draw?

For most markers yes, with potassium and chloride as the two exceptions a published comparison singles out. A review and meta-analysis of twenty-six studies found a strong overall correlation between self-collected capillary blood and venous blood. Where the answer is a yes or a no, the agreement was close to perfect.⁹

That is one pooled figure across very different markers rather than a result for each of them. The same work also found capillary collection less painful than a vein draw, and upper-arm devices less painful than a finger prick.⁹

The marker by marker picture comes from a single study. It compared a finger prick and an upper-arm device against a vein draw in 480 adults. Sixteen of the eighteen markers tested were clinically comparable under the study's own sample acceptance rules. Potassium and chloride showed limited interchangeability.¹⁰

The same study found more haemolysis with the finger prick than with the upper-arm device.¹⁰ Potassium and chloride both change with haemolysis,⁸ which we read as the likely reason those two capillary values did not match.

Across twenty-six studies the correlation between self-collected capillary blood and venous blood was strong, and agreement on yes or no answers was close to perfect. In a study of four hundred and eighty adults, sixteen of eighteen markers were clinically comparable, but potassium and chloride were not. Sources: Schroeder and colleagues, Clinical Biochemistry 2025; Deza and colleagues, Clinical Chemistry and Laboratory Medicine 2026.

Where this leaves a Diagnostics as a Service product

The honest position on self-collection is narrow rather than general. Self-collection works for most of a routine panel, and potassium and chloride are the two markers that failed. That finding comes from one comparison in 480 adults rather than from a body of work. Any supplier's own list should be just as specific about which markers it leaves out.

What does ISO 20658 require?

Less than supplier material tends to assume. The standard covers the collection and transport of samples for medical laboratory examinations, and it was published on 19 May 2023. On the text of the standard, it sets no numeric time or temperature limits. Instead it requires a laboratory to define its own limits and to show that they hold.¹¹

We have seen supplier material describe it as a European standard, which it is not. It is an international standard, and we could not find a European version of it to claim conformity with.¹¹

What a Diagnostics as a Service supplier should be able to tell you

Because the standard sets no numbers, the numbers have to come from the supplier. Ask a supplier for four things, and take the reason with each one. None of the four is a certificate.

• Which markers the supplier does not offer on capillary collection, and why. The published comparison already names potassium and chloride, so an answer that names neither is the one worth pressing on.
• The stability window for each marker, and the study behind it. ISO 20658 sets no numeric limits, so the number can only come from the supplier's own validation work.
• The rejection rate, measured rather than promised. It is the figure that shows how often a collection method fails in ordinary use.
• What happens when a sample fails, and who tells the customer and pays for the repeat. If there is no settled answer, that is worth knowing before anything is signed.

Where Aniva is on this

Aniva collects on a network of its own rather than through a subcontractor. Everything from the order to the result screen is bought once rather than assembled from five suppliers. Throughout, the partner's brand is the only one a customer meets.

Which markers belong on which collection method is worth settling before you build anything. Bring your marker list when you book a demo.

Notes and sources

1. Plebani M, Carraro P. Mistakes in a stat laboratory, types and frequency. Clinical Chemistry, 1997;43(8):1348-1351. 189 errors in 40,490 analyses, of which 68.2 percent were preanalytical. doi.org/10.1093/clinchem.43.8.1348
2. Carraro P, Plebani M. Errors in a stat laboratory, types and frequencies ten years later. Clinical Chemistry, 2007;53(7):1338-1342. 160 confirmed errors in 51,746 analyses, of which 61.9 percent were preanalytical. doi.org/10.1373/clinchem.2007.088344
3. Plebani M. Errors in clinical laboratories or errors in laboratory medicine? Clinical Chemistry and Laboratory Medicine, 2006;44(6):750-759. Gives the published range of 46 to 68.2 percent before analysis and 18.5 to 47 percent after it. doi.org/10.1515/CCLM.2006.123
4. Lippi G and colleagues. Preanalytical quality improvement, from dream to reality. Clinical Chemistry and Laboratory Medicine, 2011;49(7):1113-1126. The review that states nearly 60 to 70 percent. doi.org/10.1515/CCLM.2011.600
5. Caballero A and colleagues. Evaluation of eighteen quality indicators from the external quality assurance preanalytical programme of the Spanish Society of Laboratory Medicine. Advances in Laboratory Medicine, 2022;3(2):175-200. Median total rejection rate 2.182 percent of requests. doi.org/10.1515/almed-2021-0097
6. Lippi G and colleagues. Haemolysis, an overview of the leading cause of unsuitable specimens in clinical laboratories. Clinical Chemistry and Laboratory Medicine, 2008;46(6):764-772. Up to 3.3 percent of routine samples and 40 to 70 percent of unsuitable specimens. doi.org/10.1515/CCLM.2008.170
7. Ma C and colleagues. Real-world study, assessing the impact of hemolysis on 48 biochemical and immunological analytes through big data analysis. PLoS ONE, 2026;21(1):e0340265. Adjusted differences by haemolysis index against a reference group of about fifteen thousand samples per marker. doi.org/10.1371/journal.pone.0340265
8. Lippi G and colleagues. Influence of hemolysis on routine clinical chemistry testing. Clinical Chemistry and Laboratory Medicine, 2006;44(3):311-316. Clinically meaningful changes at mild or undetectable haemolysis, and the finding that the response was too varied for a reliable correction. doi.org/10.1515/CCLM.2006.054
9. Schröder D and colleagues. Comparison of laboratory results and pain perception in self-sampled capillary blood versus venous blood sampling, a systematic review and meta-analysis. Clinical Biochemistry, 2025;138:110965. Twenty-six studies, overall correlation 0.89, agreement on dichotomous outcomes 0.99, and lower pain with upper-arm devices than with a finger prick. doi.org/10.1016/j.clinbiochem.2025.110965
10. Deza S and colleagues. Venous-capillary blood agreement for routine chemistry analytes, effects of collection site and sample matrix. Clinical Chemistry and Laboratory Medicine, 2026. 480 adults, sixteen of eighteen markers clinically comparable, with potassium and chloride showing limited interchangeability. doi.org/10.1515/cclm-2026-0671
11. International Organization for Standardization, ISO 20658:2023, requirements for the collection and transport of samples for medical laboratory examinations, published 19 May 2023. It sets requirements for defining and validating stability and transport conditions rather than numeric limits. It is an international rather than a European standard. www.iso.org/standard/80035.html

This article describes published research on how blood samples fail before testing. It is general information rather than clinical or laboratory advice, and the choice of collection method for any given marker is a question for your own laboratory.