7.3.6.4 Data extraction

This section in the review report should include details of the types of data extracted from the included studies. Standardized data extraction tools allow the extraction of the same types of data across the included studies and are required for JBI systematic reviews. The protocol should detail what data the reviewers plan to extract from the included studies and the data extraction tool should be appended to the protocol.

The data extracted should include specific details about the participants, exposure of interest and outcomes of significance to the review question. Irrespective of the focus of the systematic review, additional data should be extracted, such as study methods, covariates and the sample size for each study included in the review. The methods of collection of exposure and outcome data (i.e. number of cigarettes or ppm of asbestos fibres or dust), which commonly include questionnaires, registries or interviews should also be stated.

Relative risk and other measures of association should be extracted, preferably those adjusted for the maximum number of covariates. Unadjusted results should be included only where no other data is provided. Epidemiological studies investigating the same association between an exposure and disease/condition provide different effect measures that may be too dissimilar to combine, which presents a challenge when combining studies in a meta-analysis. Each different study may report different measures of association, or estimates of effect, which most commonly include relative risks (RR), odds ratios (OR), hazard ratios (HR), standardized incidence ratios (SIR) or a standardized mortality ratios (SMR). An absolute risk reflects the observed or calculated probability of an outcome (disease) in a population exposed to a specific risk factor. A relative risk, which is the most common metric of risk, is simply the ratio of absolute risk in the group exposed to the risk factor of interest, to the absolute risk in a group (control) that is not exposed to the risk factor. An OR uses the odds of developing a disease in both groups to calculate a relative measure between two groups rather than the risk.

Where an absolute risk of the exposed group is presented relative to available existing data for a population group, this is referred to as a standardized ratio. Depending on whether incidence or mortality data is used will depend on whether the SIR or SMR is reported. Standardized mortality ratio refers to the ratio of observed and expected mortality, based on the age-sex-calendar period specific rates. Usually SMR greater than 1 implies higher than expected deaths and SMR less than 1 implies lower than expected deaths. Standardized incidence ratio is the ratio of the observed number of cases to the expected number of cases, based on the age-sex specific rates. A range of corrections, transformations and assumptions can be used to account for difference in the different types of data presented. 

The following details are suggested at a minimum for extraction. 

Study details
Author – this is an alphabetic or character code which is usually the first few characters of the primary study author's name. This serves as an easy way to identify the study in the bibliography
Year – the year of publication
Journal – the journal in which the article was published

Study method/characteristics
Study design – briefly describing the type of study design. For e.g. if it is a cohort study or a cross-sectional study.
Setting – may refer to hospital or community. May also refer to rural/urban etc.
Participants – includes age, sex, country/location, sample size, diagnosis and other relevant characteristics
Recruitment procedures utilized
Follow-up or study duration – any details on the duration of the study or follow-up of the participants
Exposure(s) of interest (Independent variable) – type, frequency, intensity, duration

Dependent variable (outcome) 
Outcomes – the primary outcome measured and where relevant includes associated secondary outcomes.
Outcome measurements – describe the scales or tools used to measure the outcomes, e.g. a standardized pain scale to measure pain.

Data analysis methods including statistical technique (e.g. regression), adjustment for confounding factors, etc.
Study results
Appropriate measures for effect size such as:

• Risk ratio

• Relative risk ratio

• Odds ratio

P value & 95% Confidence Intervals

Reviewer comments