| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters |
1 Department of Clinical Chemistry, University of Helsinki, FIN-00029 HUS Helsinki, Finland
2 HUCH Laboratory Diagnostics, Helsinki University Central Hospital, FIN-00029 HUS Helsinki, Finland
aAddress correspondence to this author at: HUCH Laboratory Diagnostics, Helsinki University Central Hospital, PO Box 340, FIN-00029 HUS, Finland. Fax 358-9-471-75656; e-mail janne.suvisaari{at}helsinki.fi.
To the Editor:
Reporting the result of one patient for another as a consequence of specimen mix-up, mislabeled specimens, or misidentification of patients is a serious laboratory mistake with potentially catastrophic consequences. For brevity, all mistakes in specimen identification are hereafter referred to as "specimen mix-up". Because it is assumed to be relatively frequent, specimen mix-up is usually considered one of the most important laboratory blunders. Therefore, much has been done to try to prevent it, and several result verification methods, such as delta check (1) and multivariate delta check (2), have been developed to detect it.
However, most of the available information on the frequency of this problem is anecdotal. To our knowledge, no study has quantitatively measured the frequency of all mistakes in specimen identification in a large clinical chemistry laboratory. Furthermore, the results could probably not be generalized to other laboratories. To determine how often specimen mix-up occurs in the core laboratory of Helsinki University Central Hospital, we developed a simple but accurate method to estimate the frequency of specimen mix-up and to test the method in practice. Although the results of our study may not be generalizable, our method can be used by anyone who is interested in checking the frequency of specimen mix-up in a large clinical chemistry laboratory.
Our method consists of determining the ABO and RhD blood groups, hereafter referred to as "blood group", of blood samples taken for other purposes and comparing the results with the previously known blood groups of each patient. The method assumes that the blood groups are known before the study, but this is not a major limitation because in many hospitals these groups are routinely determined in all patients, or at least in all patients in surgical wards. The blood groups of the patients to be studied could also be ascertained in the laboratory, independent of the second part of the study, the day before the study.
Not all cases of specimen mix-up will produce a discrepancy between the previously known blood group and the result of the new blood group determination, but the proportion of mix-ups that will produce a discrepancy can be estimated without a significant bias if the study population is homogeneous and the frequencies of blood groups in the study population are known. Specimen mix-up can be assumed to be a random process, or at least a process that is independent of the blood groups of the mixed-up specimens. Hence, the probability of detecting a mix-up is equal to the probability that two randomly selected samples from the study population will be of a different blood group.
We collected all blood samples taken for complete blood counts from selected wards during periods of 1 to 4 weeks. From this set of samples, we excluded the samples from those patients whose blood groups were not known beforehand. A total of 504 samples were included in the study. A routine blood group determination was then performed on these samples by laboratory technicians unaware of the previously known blood group. In addition, all other laboratory technicians and other persons who handled the samples were kept unaware of this study to prevent any bias. The ethics committee of our hospital approved the study.
The distribution of blood groups in our sample was very similar to that in the general Finnish population. The distributions did not differ in a statistically significant way (
2 test, P = 0.283). Hence, we used in our calculations the frequencies of blood groups in the Finnish population. The frequencies of blood groups A pos, A neg, B pos, B neg, O pos, O neg, AB pos, and AB neg (pos, RhD-positive; neg, RhD-negative) are 38%, 6%, 15%, 2%, 27%, 4%, 7%, and 1%, respectively. For a specimen in each group, the probability that the blood group of another specimen will be different is 1 minus the frequency of that group. Therefore, the probability of discrepancy for all groups combined is the sum of the products of each frequency and 1 minus that frequency. On the basis of the frequencies of blood groups in the general Finnish population, the probability of discrepancy is 0.7496; for the frequencies of blood groups in our sample, the result was nearly the same, 0.7493. The observed frequency of mix-ups would be the product of the probability of discrepancy and the true frequency of mix-ups. Hence, an unbiased estimate of the true frequency of mix-ups would be the observed frequency divided by this probability.
We did not detect any cases of discrepancy in blood group results. Therefore, we could not get a point estimate of the frequency of mix-ups in our hospital. We could, however, calculate the upper limit of a 95% confidence interval for the frequency of mix-ups. We first corrected n (n = number of patients) for the fact that our approach would reveal a mix-up with a probability of 0.7496. Therefore, our sample of 504 was equivalent to 504 x 0.7496 = 378.
The probability of not detecting a mix-up in n cases is (1 - Pmix)n, where Pmix is the probability of a mix-up. To obtain the upper limit of the confidence interval, we set this probability to 1 - 95% = 0.05 and obtained the following equation:
 |
 |
 |
For n = 378, Pmix = 0.00789 
0.79%. Hence, the 95% confidence interval for the frequency of specimen mix-ups in our hospital is 0.00–0.79%.
It is reasonable to assume that mix-ups do occur in our laboratory and that our sample size was too small to detect any. Whether it is worthwhile to collect a larger sample depends on how exactly we want to measure the frequency of mix-ups. If specimen mix-ups are observed, the 95% confidence interval for their frequency is the 95% confidence interval of the binomial distribution corresponding to the observed number of events (mix-ups) and the corrected sample size. The sample size is corrected by multiplying it by the probability of discrepancy obtained from the frequencies of blood groups in the patient population. The probability of discrepancy can be easily calculated, using our method, for any population and for any combination of mutually exclusive blood groups, not necessarily just the ABO and RhD blood groups. Two factors contribute to the probability of discrepancy and thus to the power of this method: the power will be higher if the number of blood groups is higher or if the frequencies are more similar.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  J W Dorigo-Zetsma, D Belewu, H Meless, E Sanders, R A Coutinho, A Schaap, and D Wolday Performance of routine syphilis serology in the Ethiopian cohort on HIV/AIDS Sex Transm Inf, April 1, 2004; 80(2): 96 - 99. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
