Data Extraction in Meta-analysis
Data extraction is an important process whereby data identified by a systematic review are extracted and prepared for meta-analysis. This is often not straightforward. This resource is designed to help you make sense of it all and avoid some common pitfalls.
Introduction
Extracting diagnostic accuracy outcomes
What can I do if the 2×2 table in a diagnostic accuracy study is not reported?
What if the prevalence is also not reported?
What if the sensitivity or specificity is also not reported?
What if neither sensitivity nor specificity are reported?
Extracting dichotomous outcomes
What can you do when prognostic studies report measures of risk on different scales?
What can you do when a study reports a beta coefficient instead of a hazard ratio?
How do you calculate a standard error of a beta coefficient?
How can categorical risk data be pooled?
A worked example using a trend estimation method to summarise categorical risk data
Wanting a particular reference category in categorical risk data
What if something is missing from categorical risk data?
Estimating a hazard ratio from time-to-event data
Estimating a hazard ratio from a Kaplan curve and information about follow-up
Estimating a hazard ratio from a Kaplan curve and numbers at risk
Extracting continuous outcomes
What should I do if I have a missing mean, standard deviation or sample size?
What if the data I want are reported for the wrong time point?
What if the summary statistic I want is given for the wrong group?
Obtaining summary statistics by using complementary equations
What if you’re missing a mean and only a similar statistical statistic is given?
What if you’re missing a standard deviation and only a similar summary statistic is given?
What if neither the summary statistic I want nor a similar statistic are reported?
What if the study only reports an effect estimate?
Other approaches to dealing with diverse continuous outcome data
General issues
Logarithms and log-transformations
How can you make data extraction more efficient?
How can you reduce the risk of errors and bias when extracting data?
Extracting graphical data in fewer words
pico, PICO (No, I’m not shouting)