From: Assessing variability in results in systematic reviews of diagnostic studies
Method | Number | Percent |
---|---|---|
Graphical | ||
Plots present in article | ||
No plots | 5 | 9 % |
Forest Plots | 34 | 64 % |
ROC | 40 | 75 % |
Both | 26 | 49 % |
Statistical | ||
Cochran’s Q test | 28 | 53 % |
I2 | 31 | 58 % |
Confidence Intervals presented | 7 | 13 % |
τ2 | 7 | 13 % |
From a univariate analysis | 6 | 11 % |
From a bivariate analysis | 1 | 2 % |
Prediction intervals, ellipses, or bands | 3 | 6 % |
Provided a definition for significant variability | 24 | 49 % |
Cochran’s Q test | 10 | 19 % |
I2 | 7 | 11 % |
Cochran’s Q test or I2 | 7 | 13 % |
Influence of variability on analysis approach (reported by authors) | ||
Whether to perform a meta-analysis in the first placea | 1a | 4 %a |
Whether to use a fixed or a random effects model | 16 | 30 % |
Whether to investigate sources of heterogeneity | 4 | 8 % |
How variability in results is mentioned in the abstract and discussion and/or conclusions | Discussion/Conclusions | Abstract |
Any mention of variability listed below | 29 (55 %) | 15 (28 %) |
A vague discussion variabilityb | 17 (32 %) | 10 (19 %) |
Reported results of a statistical test or measurement of variability | N/A | 4 (8 %) |
Variability in results precludes firm conclusions or is a study limitation | 13 (25 %) | 2 (4 %) |
Despite variability in results, a conclusion could still be made | 7 (13 %) | 3 (6 %) |
There was no relevant variability in results | 1 (2 %) | 2 (4 %) |