Introduction

With the improvement of laboratory management practices, increased state accreditation, and the growing popularity of national laboratory accreditation in China's medicine and health field, the process of uncertainty evaluation in medicine and health laboratories has gained more attention and recognition. The promotion of uncertainty evaluation in China has almost reached its 20th anniversary, from the guidance of GUM:1995 [1] to JJF 1059–1999 [2], JJF 1059.1–2012 [3], and the implementation of QUAM 3rd (CNAS GL06) [4, 5] and RB/T151-2016 [6] for uncertainty evaluation guidelines in chemical measurement and microbial quantitative analysis. The development of uncertainty evaluation in medicine and health is not as rapid and widespread as in the fields of physics and calibration laboratories, but in recent years, it has rapidly progressed and has become a routine work required for laboratory accreditation such as CNAS and CMA. However, understanding the rationality of uncertainty evaluation has always been challenging, leading to numerous errors in daily evaluations. This makes it difficult for frontline personnel to objectively assess uncertainty and obtain reliable and accurate results. In addition, there is an urgent lack of literature on uncertainty error analysis for reference among peers. Therefore, this paper aims to analyze the common errors in uncertainty evaluation in the field of medicine and health and propose correct methods for evaluating measurement uncertainty, using literature research and international standard comparison.

Methods

Literature research and measurement methods

The China National Knowledge Infrastructure (CNKI) literature database was used for uncertainty literature research and measurement, using "uncertainty" as the keyword and “medicine and health” as the searching condition. The research period was not limited, and the temporal sequence of academic attention was analyzed using the CNKI literature index. As there were relatively few uncertainty literature before 2000, this research divided the journal literature into five periods: before 2000, 2001–2005, 2006–2010, 2011–2015, and 2017 to present, using the number of journal literature during these periods as the stratified sampling basis.

Literature sampling scheme

A total of 30 journal articles were selected as representatives for uncertainty error analysis and statistics, using a stratified proportional sampling method to sample the CNKI uncertainty assessment journal literature in the field of medicine and health.

Classification and basis of uncertainty errors

By comparing QUAM 3rd and CNAS GL06 (the Chinese translation of QUAM 3rd), seven types of errors were identified which included: E1. Incomplete mathematical models; E2. Errors in the standard curve formula evaluation; E3. Inappropriate standard series evaluation; E4. Insufficient information for type B evaluation; E5. Uncertainty of the blank not evaluated; E6. Improper recovery rate evaluation; E7. Improper significant figures. The qualitative description and explanation of these errors are shown in Table 1.

Table 1 Common errors in uncertainty evaluation and characterization
Full size table

Statistical methods

Rate comparison was performed using the chi-square test, which was conducted using IBM SPSS Statistics 19.

Results and discussion

Uncertainty academic attention and journal literature statistics

The keyword uncertainty was selected from the CNKI literature database, with no time limit, and a total of 28,911 documents were retrieved. The literature on uncertainty in China has grown significantly since 2000 and peaked in 2013, then gradually declined. Foreign literature also shows a similar trend, but after 2013, there was a sharp increase in foreign literature, which may be related to the increasing attention to uncertainty in QUAM 3rd and the medicine and health field. The number of journal articles retrieved with “medicine and health” as the search restriction was 2568. The statistics of the publication quantity of Chinese journal articles are shown in Fig. 1, which increased significantly after 2005 and reached its peak in 2011–2015. This may be related to compulsive national accreditation and laboratory accreditation requirements.

Fig. 1
figure 1

Trends in the number of publications of Chinese uncertainty journals in the Chinese National Knowledge Infrastructure (CNKI) literature database

Full size image

Uncertainty error statistical analysis

Using the stratified proportional sampling method, 30 journal articles were randomly selected from the 2568 journal articles in the CNKI literature database, and the uncertainty error classification and statistics were carried out. The number of samples for each of the five periods, including before 2000, 2001–2005, 2006–2010, 2011–2015, and 2017 to present, was 0, 2, 9, 15, and 4, respectively. The articles were from 15 provinces and municipalities, including food and drug, health systems, and universities at the national, provincial, and lower levels. Among them, 11 articles used atomic spectroscopy methods including ICP/MS, 12 articles used chromatography methods including chromatography-mass spectrometry, 5 articles used photometric methods, and 1 article each used titration and enzyme-linked immunosorbent assay methods. The error rate statistics are shown in Table 2. None of the 30 journal articles were error-free, and the total error rate was as high as 44 %. The four types of errors with higher error rates include: blank uncertainty without evaluation error rate of 87 %, improper standard series evaluation error rate of 67 %, inappropriate significant figures error rate of 60 %, and insufficient information of type B evaluation error rate of 50 %. The error rate of provincial and above institutions was 48 %, while the error rate of institutions below the provincial level was 43 %. The difference in error rates between the two was not statistically significant with p = 0.523. This partially reflected that provincial and above institutions have not played a good technical guidance role in the correctness and rationality of uncertainty evaluation methods.

Table 2 Statistics of common error rates in uncertainty evaluation of 30 journal articles
Full size table

Conclusion

The academic attention to uncertainty in the CNKI literature database showed explosive growth after 2005, and the number of journal articles retrieved with "medicine and health" as the search restriction showed a similar growth trend. However, the total error rate of the 30 randomly sampled journal articles, compared with the seven types of errors summarized in QUAM 3rd and CNAS GL06 recommended terms, was as high as 44 %, and no journal article was error-free, indicating that the correctness and rationality of uncertainty evaluation in the medicine and health field are not optimistic even terrible. The difference in error rates between provincial and above units and units below the provincial level was not statistically significant (p = 0.523), partially reflecting that provincial and above institutions have not played a good technical guidance role in the correctness and rationality of uncertainty evaluation methods. The lack of education on measurement uncertainty in the medicine and health field, the quality of continuing education, and the lack of academic communication of uncertainty evaluation shows the pain points of the current severe situation. Improving the quality of uncertainty education and communication both multi-dimensionally and multi-facetedly should become an urgent need in the medicine and health field.

The current study may be subject to at least three possible limitations. First, due to the small sample size problem, there may be sampling error from literature research. Second, the keyword “medicine and health” was limited by Chinese expression, the published articles only indirectly reflected the errors in daily practices, but our results provided valuable reference meanings for present severe situation of measurement uncertainty in medicine and health field. Finally, as the current study did not involve the uncertainty evaluation of the microbiological field, a status assessment of this field could not be performed.