this is for holding javascript data
Matt Vassar edited textbf_Results_textbf_Literature_search__.tex
almost 9 years ago
Commit id: 9d5cadf358463d05b63adc6a192ab3f42e06d0d3
deletions | additions
diff --git a/textbf_Results_textbf_Literature_search__.tex b/textbf_Results_textbf_Literature_search__.tex
index d8ca726..59b0b27 100644
--- a/textbf_Results_textbf_Literature_search__.tex
+++ b/textbf_Results_textbf_Literature_search__.tex
...
\textbf{Results}
\textbf{Literature search }
The Pubmed search resulted in 337 articles from four journals. After
Covidence screening
for systematic reviews and/or meta-analyses titles and abstracts via Covidence, 79 were
exclude search. Coder consensus excluded excluded. Full-text article screening resulted in removal of an additional 74 articles which
did not met the definition of
a systematic review and/or
meta-analysis, yet were better categorized into other study types (Individual patient data study, human trials, Histologic studies, Genetic studies, letters to the editor, and Genomic microarray meta analyses). meta-analysis. Additionally 2 studies could not be retrieved. In total 182 manuscripts were analyzed for heterogeneity (Figure 1).
\textbf{Stata 13.1 Heterogeneity assessment}
Fifty percent (91/182) of all
manuscripts analyzed utilized some form or combination of quantitative meta-analyses used at least one heterogeneity test (Figure 2).
A The most widely reported statistic was I2 (9.63\%; 21/218) followed by X2 (8.26\%; 18/218). In combination, X2 and I2 (12.84\%; 28/218) were reported with greatest frequency followed by Q and I2 (12.39\%; 27/218). Authors selected a random-effects model
more frequently than was the most common meta-analysis used 23\%(42), next came both fixed and random-effect models 18\%(33), fixed-effects was used in 7\%(13) of the manuscripts, and lastly a mixed-effects model was used in 0.92\%(2) of the available manuscripts (Figure 3). This leaves roughly 50\%(91) of available manuscripts with no meta-analysis model in use to quantitatively determine heterogeneity (Figure 3). Of the 42\%(76) which used the random-effects model, 23\%(18) used that model without performing a heterogeneity test to confirm the need for a random-effects model. 12\%(22) of the manuscripts changed from the fixed-effects meta-analysis model to the random-effects after a heterogeneity test. The significance level most used was a 5\% p-value with a 61\%(111) occurrence and 0.1\%, 1\%, 20\% as the least frequently used with 2\%(4), 4\%(8), and 2\%(4) occurrences respectively. A forest plot was the most used heterogeneity plot with 40\%(73) while 1\%(2) used a forest and a L’Abbe heterogeneity plot. Out of the manuscripts which created a heterogeneity plot 41\%(37) actually presented them. Of the three tests designed to investigate heterogeneity (Subgroup, Meta-regression, and Sensitivity Analyses), Subgroup analysis was used the most 22\%(40), Sensitivity analysis was second 18\%(33), and Meta-regression was used the least 8\%(15) (Figure 4). It was found that 19\%() of the available manuscripts wrote about heterogeneity, but never actually calculated it. 57\%(104) of manuscripts did not find significant heterogeneity, 2\%(4) found enough evidence of heterogeneity to disregard “some” of the meta-analysis, 4\%(7) found significant heterogeneity, and 37\%(67) did not know whether or not there was too much heterogeneity present to perform a meta-analysis.Summarization of evidence mapping efforts
\textbf{First Objective}
Table 1 from the evidence mapping method from (Althius 2014) was applied to a systematic review, Maltoni et al. This qualitative method of me . (not done with this one yet, i'm figuring out how to format it in excell)
...