Skip navigation

Blunders


A dictionary definition of a blunder is "a gross mistake". As a verb it is also defined as "to mismanage". Because we live in an imperfect world, we expect blunders to occur. How we feel about them depends - if we make them they are less of blunder than if someone else makes them. If the battery in our new car fails on trip to the supermarket we might be less cross than if it fails in the middle of torrential rain in the middle of one of those empty bits of Wales or Scotland or on a bad day on the M25.

But most of us would accept that while blunders do occur, we should take every reasonable step to see that they do not. Bandolier was speaking with a group of GPs about diagnostic tests, and was surprised to learn that none of them were aware of blunder rates in tests, which in the past have been said to occur at about 1%. So we thought we would see what the literature said about blunders, and found some interesting papers in clinical biochemistry.

Defining a blunder

You know one when you see one, so some true-life examples:

  • A high drug level is reported in Mrs X. But Mrs X isn't taking that drug and didn't have a blood sample taken! It is discovered that actually the sample was from Mr Y, but that the wrong labels were used on sample and request form.
  • An imaging report says there is an abdominal mass in a patient sent for a head and neck scan (wrong report).
  • Five patients are all reported to have very high prolactin levels. It turns out that they were all analysed together, and that the first had so much prolactin that it resulted in significant carry-over into the next four specimens.
  • A new computer system is installed which fails to recognise results with three integers - only two. So a result of 123 is reported as 23.
These examples point out the areas where blunders can occur - in ordering tests, in the analysis stage and in transcribing and mis-reporting errors. Errors may be picked up by the laboratory, or those receiving the results, and, importantly, by external quality control schemes.

Analytical errors

A study of 220,000 individual clinical chemistry results obtained in methods comparison studies in the USA compared each result with its replicate [1]. It found 98 examples where replicates were >=7 standard deviations from the expected value and an additional 360 in which the difference was 4 to 6 standard deviations from the expected result.

This gives a crude error rate of 0.045% at the 7 SD level and 0.081% at the 4 SD level. That is, less than 1 analytical blunder per 1000 analyses. But of these, only nine results were sufficiently different to be judged as likely to influence patient care.

An influential Scottish study [2] came to similar conclusions for analytical blunders. In a study over three time periods involving almost 300,000 results the analytical blunder rate was 0.034%. But the total blunder rate, including ward, sample handling, reporting and clerical errors produced a total blunder rate 10 times higher at 0.3%. The blunder rate for external quality assessment samples was 0.2%.

Similar results (an overall rate of about 0.3% of results) resulted from an English study in two clinical biochemistry laboratories over one year involving 248,000 samples and 997,000 results [3]. The two laboratories had blunder rates for external quality assessment samples of 0.5% and 0.2% per analyte and 1.7% and 0.7% per sample respectively [4].

An Australian perspective

In 1994 14 laboratories from all over Australia participated in a study to examine laboratory errors [5]. Each randomly selected 100 hand-written pathology request forms, and they scored the number of transcription errors, defined as any instance where the data on individual request forms were not identical to the data entered into the laboratory's computer system. Laboratories also scored the total number of quality assurance samples analysed and the proportion of results that lay outside the allowable limits of error of the programme.

Transcription errors varied widely. While most laboratories clustered around median error rates of 1% to 3%, at least two laboratories were more than 2 standard deviations above the mean error rate, as the box shows.

Transcription error rates on pathology request forms [5]
 
Error rate (%)
Details Best laboratory Median laboratory Worst laboratory
Patient identification 0 1.0 9
Patient sex and age 0 2.0 17
Patient ward or address 0 2.5 9
Tests requested 0 2.5 15
Requesting doctor identification 0 1.5 17
Analytical errors from one cycle of analysis in 1993 and 1994 occurred in up to 26% of analytical results, and were above 10% in eight of the 14 laboratories.

The combined error rates (transcription plus analytical) were calculated as being up to 46% in the worst performing laboratory. One laboratory had more than 95% of error-free results, while six of the 14 laboratories had better than 80% of error free results.

The primary care perspective

All of the above, of course, examines the problem of blunders in laboratory testing principally from the laboratory perspective. But what about the consumers - do they see blunders? In the USA 124 primary care physicians in 49 practices participated in a prospective study in which they reported problems over a six month period [6]

They reported 180 problems, of which just over a quarter were judged to have an effect on patient care. The crude result was 1.1 problems per 1000 patient visits (or about 0.1%). But not all patient visits result in a sample being taken for analysis. A good estimate was that about one-third of all visits to these practices resulted in a blood sample being taken for laboratory analysis. So the best estimate of problems with laboratory testing in primary care in the USA is 3.4 per 1000 visits, or 0.34% - a figure remarkably similar to these found in the two UK laboratory studies. Analytical errors were only about 10% of the total, at 0.044%.

Comment

Lest any of Bandolier's clinical biochemistry friends think that we are picking on them specifically, it is worth pointing out that clinical biochemistry has long been in the van of quality assessment and quality improvement. That is why we have so much data!

There is some consistency in the findings. Overall, the rate of blunders seems to be pretty constant at about 0.3% of results (3 per 1000), and the analytical error is perhaps 0.04% (less than 1 per 1000). So blunders do happen in the best regulated systems, so if a result looks wrong, it is probably worth checking. And there will be blunders that never get picked up. If Mrs A's sample for thyroxine is mixed up with Mrs B's, and both are normal, who could tell?

Summary of results
Blunder rates (% of results or samples)
Reference Analytical Overall Comment
White et al [1] 0.05 n/a Results
Chambers et al [2] 0.04 0.30 Results
Lapworth & Teal [3,4] 0.35 0.30 Analytical on QA samples
Khoury et al [5] 11.40 n/a QA samples outside limits
Nutting et al [6] 0.04 0.34 Samples
The Khoury study [5] sticks out like a sore thumb, and it is hard to understand how such high rates of analytical error are allowed in quality assurance schemes (though perhaps Bandolier is missing something). Perhaps, as the paper comments, the fact that there is no minimum standard of performance which laboratories are required to maintain is an important negative influence on overall performance. In the UK there are (or perhaps were is a better description because this may be under threat) mechanisms whereby poorly performing laboratories were given help and advice. This meant that no laboratory could perform poorly on one test, or on many tests, consistently.

Even so there is a clear message from this important study. In the same circumstances some laboratories do brilliantly, while some are awful. The equipment, the funding, and people probably don't differ by much. But management and leadership can make a huge difference to quality of service.

References:

  1. DL Witte, SA VanNess, DS Angstadt, BJ Pennell. Errors, mistakes, blunders, outliers, or unacceptable results: how many? Clinical Chemistry 1997 43: 1352-6.
  2. AM Chambers, J Elder, D O'Reilly. The blunder-rate in a clinical biochemistry service. Annals of Clinical Biochemistry 1986 23: 470-3.
  3. R Lapworth, TK Teal. Laboratory blunders revisited. Annals of Clinical Biochemistry 1994 31: 391.
  4. R Lapworth, TK Teal. Laboratory blunders revisited. Annals of Clinical Biochemistry 1994 31: 78-84.
  5. M Khoury, L Burnett, MA Mackay. Error rates in Australian chemical pathology laboratories. Medical Journal of Australia 1996 165: 128-130.
  6. PA Nutting, DS Main, PM Fischer et al. Problems in laboratory testing in primary care. JAMA 1996 275: 635-9.



previous or next story in this issue