Table 2. Percentages of errors in quotations of scientific articles in medicine.
Main analysis and sensitivity analyses. Estimates are based on references (the denominator is the number of references in an article) or on quotations (the denominator is the number of quotations in a text. According to studies from this systematic review (Buijze, Davids, George, Gldberg, Lukic, Luo, Neihouse), the average reference is quoted about 1.7 times in an article (range: 1.37–1.99). Estimates based on references can differ in the number of errors counted per reference: only 1 or >1 (“reference based, restricted” vs. “reference based, unrestricted”). Main analysis is based on all studies. If more than one approach was reported the default was >1 error counted on the basis of references (reference based, unrestricted). Random effects models were employed in all analyses. Indirect references: references to a secondary source, such as a review article instead of the original article. Low vs. high risk of bias analysis is based on main analysis. Total N: 7,171 references. 95% confidence intervals in square brackets, number of studies, and I2-statistic as measurement of heterogeneity.
| Main analysis | Sensitivity analyses | |||||
|---|---|---|---|---|---|---|
| Reference based, restricted | Reference based, unrestricted | Quotation based | Without indirect references | Low vs. high risk of bias | ||
| Major errors, % [95% CI] Studies, I2-statistic | 11.9 [8.4, 16.6] 27 studies, I2: 95% | 12.3 [9.3, 6.1] 15 studies, I2: 82% | 11.6 [6.1, 20.8] 12 studies, I2: 98% | 11.9 [6.9, 19.9] 6 studies, I2: 97% | not applicablea | 12.6 [7.1, 21.3] vs. 11.3 [8.8, 14.3] p = 0.713 |
| Minor errors, % | 11.5 [8.3, 15.7] 27 studies, I2: 95% | 10.6 [6.4, 17.3] 15 studies, I2: 94% | 12.2 [8.0, 18.3] 12 studies, I2: 95% | 10.0 [4.8, 19.6] 6 studies, I2: 98% | 8.5 [6.8, 10.7] 27 studies, I2: 84% | 12.6 [7.8, 19.6] vs. 10.6 [7.0, 15.7] p = 0.585 |
| Total errors , % | 25.4 [19.5, 32.4] 28 studies, I2: 97% | 24.8 [17.3, 34.3] 15 studies, I2: 94% | 26.1 [17.4, 37.2] 13 studies, I2: 98% | 21.8 [10.5, 40.0] 6 studies, I2: 98% | 21.5 [17.4, 26.2] 28 studies, I2: 94% | 29.9 [19.5, 42.8] vs. 21.5 [16.4, 27.7] p = 0.189 |
Notes.
In quotation accuracy studies, indirect references are invariably counted as minor errors. Therefore, subtracting secondary references from the sum of errors (where possible) does not change the figures for major errors.