When adapting or designing a data collection form, review authors should first consider how much information should be collected. Collecting too much information can lead to forms that are longer than original study reports, and can be very wasteful of time. Collection of too little information, or omission of key data, can lead to the need to return to study reports later in the review process.
Here are some tips for designing a data collection form, based on the informal collation of experiences from numerous review authors. The checklist in Table 7.3.a should also be consulted.
Include the title of the review or a unique identifier. Data collection forms are adaptable across reviews and some authors participate in multiple reviews.
Include a revision date or version number for the data collection form. Forms occasionally have to be revised, and this reduces the chances of using an outdated form by mistake.
Record the name (or ID) of the person who is completing the form.
Leave space for notes near the beginning of the form. This avoids placing notes, questions or reminders on the last page of the form where they are least likely to be noticed. Important notes may be entered into RevMan in the ‘Notes’ column of the ‘Characteristics of included studies’ table, or in the text of the review.
Include a unique study ID as well as a unique report ID. This provides a link between multiple reports of the same study. Each included study must be given a study identifier that is used in RevMan (usually comprising the last name of first author and the year of the primary reference for the study).
Include assessment (or verification) of eligibility of the study for the review near the beginning of the form. Then the early sections of the form can be used for the process of assessing eligibility. Reasons for exclusion of a study can readily be deduced from such assessments. For example, if only truly randomized trials are eligible, a query on the data collection form might be: ‘Randomized? Yes, No, Unclear’. If a study used alternate allocation, the answer to the query is ‘No’, and this information may be entered into the ‘Characteristics of excluded studies’ table as the reason for exclusion.
Record the source of each key piece of information collected, including where it was found in a report (this can be done by highlighting the data in hard copy, for example) or if information was obtained from unpublished sources or personal communications. Any unpublished information that is used should be coded in the same way as published information.
Use tick boxes or coded responses to save time.
Include ‘not reported’ or ‘unclear’ options alongside any ‘yes’ or ‘no’ responses.
Consider formatting sections for collecting results to match RevMan data tables. However, data collection forms should incorporate sufficient flexibility to allow for variation in how data are reported. It is strongly recommended that outcome data be collected in the format in which they were reported (and then transformed in a subsequent step).
Always collect sample sizes when collecting outcome data, in addition to collecting initial (e.g. randomized) numbers. There may be different sample sizes for different outcomes because of attrition or exclusions.
Leave plenty of space for notes.