(Image from https://blog.webhose.io/2015/12/08/how-to-extract-data-from-a-website-5-steps-to-transform-unstructured-data-into-business-insight/)

Data extraction is the process of collecting information reported by the included primary studies, such as:

• Participant characteristics (e.g., age, gender, ethnicity, socio-demographic details, comorbidities)
• Setting (e.g., pre-hospital vs. hospital, geographical location, dates)
• Diagnostic criteria (e.g., definition of heart failure or hypertension)
• Intervention information (e.g., number and length of interventions; route, dose, or timing of delivery)
• Outcomes of interest (e.g., summary statistics)
• Study metadata (e.g., authors, publication date, funding source, conflicts of interest)

What you will need:

1. Data extraction form: a paper or electronic form used to collect desired information from primary studies. Because each review attempts to answer a different question, each data extraction form will also be different. 

2. Two independent reviewers: at a minimum, information that involves subjective interpretation or is critical to the interpretation of results (e.g., outcomes) should be independently extracted by at least two people who reach consensus in cases of disagreement.

3. Preparedness to contact the original investigators: while most data will be obtained directly from the study reports, it may be necessary to contact the original investigators in cases of incomplete reporting. 

(Image from http://www.robinkok.eu/2016/03/11/risk-of-bias-assessment-graph-in-excel/)

Systematic reviews usually include a risk of bias (ROB) assessment. A bias is a systematic error in the design, conduct, or reporting of a study that can lead to underestimation or overestimation of the true effect size, including:

• Selection bias
• Performance bias
• Detection bias
• Attrition bias
• Reporting bias

What you will need:

1. ROB assessment tool: choose one or more ROB assessment tools depending on your systematic review question. 

2. Two independent reviewers: at a minimum, ROB assessment should be independently performed by at least two people who reach consensus in cases of disagreement.

3. Presentation of results: ROB assessment results are usually presented as a ROB graph or a ROB summary figure.

(Image from https://communities.sas.com/t5/SAS-GRAPH-and-ODS-Graphics/A-problem-of-row-space-in-the-forest-plot/td-p/268448)

Meta-analysis is the statistical combination of results from two or more primary studies, typically by calculating a weighted average of the effect estimates from different studies. 

What you will need:

1. Realization that meta-analysis is not always appropriate: a meta-analysis should not be performed in situations in which the primary studies are clinically diverse or at high risk of bias or in the presence or serious reporting or publication biases.

2. Consideration of statistical methods: critical considerations include which comparisons to make, how to calculate summary statistics (e.g., odds ratio, risk ratio, mean difference), and how to test for heterogeneity among primary studies. Need more help? Contact the WSU Clinical Research Services Center (need link) or Research Design and Analysis Unit.

3. Visualization of results: meta-analysis results are usually presented as a forest plot.