As the synthesist says: nothing in this world is certain, except death, taxes, and citations. In this two-part series, we try to bring some clarity to MetaReviewer’s approach to the latter.
Pre-MetaReviewer Stage
Citations that are potentially eligible for an evidence synthesis can and should derive from numerous sources – online databases, forward and backward reference harvesting, searches of applicable organizational website, citationchaser, emails to authors asking (begging?) for unpublished studies. An important indicator of a high-quality evidence synthesis is the lengths the synthesis team goes to find every potentially relevant citation.
Unless you are doing something wildly outside the norm, most of your citations will come from a searchable, online research repository like EBSCOhost or ProQuest. Numerous researchers, librarians, and application help desks’ have written about the process of downloading citations from these databases. We won’t walk through all the processes here. The result of these processes – including the conversion from an RIS to CSV file – will result in something that should look like the below picture.
Pro Tip: EBSCOhost and ProQuest functionalities vary by institution. This is especially true when it comes to bulk downloading of citations. I learned this lesson a few years ago. I was teaching an introductory course on systematic reviewing, discussing this very topic. Because I was naïve (or perhaps, forgivingly, just inexperienced?) I assumed that everyone’s EBSCOhost looked and acted the same as mine. It never occurred to me that – perhaps this is the naïve part! – companies operating and maintaining databases like EBSCOhost might charge institutions for varied levels of access to functionality.
I had prepared my lecture assuming that everyone could follow instructions like these and easily download all found citations. At this particular institution, where I was logged into via their Wi-Fi, I no longer had access to bulk downloading. Instead, researchers – and now me, too – at this institution had to download citations 50 at a time. Assuming 5 minutes per download, it would take a user nearly 17 hours to download 10,000 citations! I was left dumbfounded and grateful that my institution paid for the feature.
So, if your institution doesn’t download every found citation all at once, consider asking your librarian to look into upgrading their current functionality. I no longer assume everyone’s access looks like mine, too.
Preparing to Upload into MetaReviewer
You will be ready to upload citations into MetaReviewer once you’ve: (a) found and downloaded all citations that might be relevant from all sources, and (b) removed duplicates from this set of found citations. (MetaReviewer’s current capabilities do not include de-duplication. Watch this space for this future capability!)
The next step involves a bit of spreadsheet processing to fit our pre-specified format. First, download the citation template file. The picture below shows where you’ll find it within the “Add Citations” button on the left sidebar within the Project management page. Readers can also directly download the file here.
Let’s examine the Citation Template File variable names. If you look closely and compare them to the names in the CSV file that we downloaded from EBSCOhost, you will notice (purposeful) similarity. We tried to maintain a consistent variable naming and order to make it easy to format the Citation Template File.
The reason we wanted to make it easy, and this is important, is because of this simple fact: you must upload the Citation Template File to MetaReviewer in the exact same format. Meaning, the same variable names and order. Any additional columns added by a user will be completely ignored.
When I manipulate this file, I take one of two approaches. (Honestly, both works fine, depending on your particular workflow.) The non-reproducible approach is simply to copy/paste from the original file into the Template. It takes a few minutes to go through each column, but not touching the variable names ensure that they don’t change and everything is set the correct way. The other approach is to use R (or any statistical software), write a script that uses the exact format and variable names, and move the downloaded citations to the Template that way.
To end, let’s examine one important aspect of the file and how MetaReviewer processes it: the Citation ID and Study ID. Users need to create a unique Citation ID for each and every citation – so one Citation ID per row. (We provide more details on ID values in the MetaReviewer User Guide.)
The Study ID doesn’t have the same requirement. We designed MetaReviewer to understand that sometimes two or more Citation IDs represent a single Study ID. (In my own work, I’ve been involved with such a phenomenon.) In these cases, users should repeat the same Study ID, which will tell MetaReviewer to link citations together within the same study.
Looking Ahead
In the next installment of #NewYearNewReview, we will discuss how to manage citations within the MetaReviewer application. Stay tuned!