• Using the title and abstract you need to decide which articles should be included
• Decide what is eligible and is to be included and record the reasons why things were excluded
• Decide what you need to find in full text, if you cannot decide from the title and abstract what should be included
• Sometimes you need to see the full text before you can decide if something should be included or not, the abstract might not give enough detail to make a decision. Again at this stage you would give your reasons for any exclusions
• Now you need to decide what sort of analysis you would do with the results you have found. It will be either quantitative (a meta-analysis) using some statistical tools or qualitative (a narrative analysis) using a written assessment. Sometimes you need to see all the results before you can decide what sort of analysis you can do. The results you end up with are not always suitably similar enough to do a meta-analysis but you cannot know this at the outset. 

Looking at the individual studies and depending on whether they are homogeneous or heterogeneous and how comparable they are will determine what methods of analysis you can use.

Some people find Rayyan useful for screening the results. This is a free web application for managing and sorting your references.

Data Collection Form will help as you examine each study to determine if it should be included. You must answer as many of the questions in the Data Collection form as you can.  The Cochrane Library EPOC page has an example of a Good Practice Data Collection Form that can be edited to suit your requirements.

Data extraction depends on the conditions you are looking at and how it might impact your conclusion.  For example, would three different studies with a population of 30 people in South America, a population of 70 people in Canada, and 150 people in North East England actually be comparable. Could you state categorically that something that happens to people in South America is likely to happen to people living in Newcastle. If all three studies took a population internationally then they would be comparable but you need to bare these differences in mind.

Part of the data extraction process is pulling out the actual numbers of people in the studies.  Can you go with the 250 people and put that population group together and then add on all of the rest of the studies after that and apply the results across the board or will there be differences because of the way the study was carried out. Sometimes the measurements will be different, for example, if you are measuring someone's adherence to medication are you using a methodology to measure that adherence e.g. the Morisky Scale or has it been reported anecdotally by them saying "yes I took my medication". Do the two equate and can you put them together.

Outcomes can be measured differently depending on the aim of the study. For example if you were looking at the effectiveness of a drug in treating cancer patients, one study might look at quality of life but another study might look at morbidity and how long the drug kept the person alive but not consider quality of life. You need to decide if these outcomes can be compared.

You might want to consider using GRADE CERQual the aim of which is to transparently assess and describe how much confidence decision makers and other users can place in findings from qualitative evidence syntheses.

You will need to have these conversations with your tutor or supervisor the Library is not able to advise on these types of decisions.

To help with the writing up part of your systematic review take a look at other reviews published by the Cochrane Library to see how the results should be presented.

Use the same types of section headings and format - it is a tried and tested method.

You should be very clear about:

• What kind of review you have done
• Why you did this kind of review

One school of thought is that a meta-analysis is a better option for a review but it is not always possible to do this with the results you find. You need to justify what you have done. If you do a narrative analysis and a meta-analysis was possible, then one of your conclusions might be, that the next step should be to carry out a meta-analysis but say why you did not do this initially.