Version 18.104.22.168 introduces a brand new "Reconciliation" feature for EPPI-Reviewer Web, designed to support most reconciliation needs for data-extraction scenarios. Previously, it was necessary to use comparison coding reports on a per-item basis, and editing the coding directly in the "Item Details" page. The new screen allows to reconcile coding differences that apply to complex coding tools directly from the reconcile page. This release includes also other new features, enhancements and bugfixes (mostly for EPPI-Reviewer Web).
New feature: reconcile complex coding
The reconcile window already allowed to reconcile coding differences on a per "comparison" way, very quickly and effectively, if dealing with screening decisions or with really simple "standard" coding tools. When the comparison applied to coding for a complex coding tool (with lots of and/or with deep branches) until now the recommendation was to "list" disagreements, produce a comparison report for the first item, use it to "adjust" the coding done by the current user and then complete it. This worked, but was time consuming, required a lot of attention and was thus prone to error.
To resolve this, the "reconcile" page now offers a new way to "see" coding differences and interact with them, accessible from a new button (top left) called "Show Detailed Tree-View".
In this "view" each reviewer in the comparison has their own column with its "version" of the coding tool, allowing you to examine each person's work in detail, for each item listed as a disagreement in the comparison.
The view also indexes and allows to browse individual "coding disagreements", which are identified at the level of a given code (and Arm, if any), but do not go as far as evaluating the contents of the Info Box and/or the coded text (full text coding). The reason for this is that you can expect that pretty much everything would be classified as a disagreement if the system evaluated these details.
It is possible to visualise the details of full text coding (on demand, via the apposite option) as well outcomes (ditto). The contents of the Info Box text are available by default, as well as coding that applies to Arms.
In data-extraction scenarios, it is not reasonable to expect that the coding done by a single reviewer will be "perfect" as is, thus, simply allowing to "complete" one version is unlikely to be sufficient. For this reason, this same view allows to Copy the coding from one reviewer to the other. Review Administrators can perform this task freely, other people can only copy other reviewer's coding onto their own version (if present).
New Feature: "All coding" (mega) report
In EPPI-Reviewer Web, review administrators can now request and save directly to their local file system a comprehensive, per coding-tool "coding report", which includes all coding data, in tabular form. This report might be useful for safekeeping, complex data analysis (it can be imported to excel), troubleshooting, and more. It is accessible from the "Review home" tab, by clicking on the coding tool name in the "Coding Progress" area. If the current user has administrator rights over the current review, a new button (Get "all coding" report for this tool) will appear at the bottom of the coding figures.
The corresponding report (which might be huge!) reports the codes as columns and items as rows. Each column header mentions the full path of a given code. In each cell, coding will be listed on a per Reviewer basis, including incomplete coding (if any), which is displayed with a different format, to distinguish it from completed coding. The report also includes a second table with "outcomes" data (if present).
New Feature: apply classifiers built in a different review
Until now, it was not possible to build a classifier in one review and then apply it in another review, unless we (EPPI Centre people) manually made this possible on a per-review, per-classifier basis. With this release, in EPPI-Reviewer Web the "classify" panel (in the "Search and classify" tab) allows to display two lists: (1) the list of classifier built within the current review and (2) the list of classifiers built from any review the current user belongs to. Naturally, it is possible to "rebuild" only classifiers from the first list.
New Feature: search by source (and more)
The "New search" panel of EPPI-Reviewer Web now contains a new option called "From Source(s)". This allows to create a search listing all items from one or more sources and can be further refined by specifying a secondary option (available values are: "All items in source", "Only included", "Only excluded", "Only deleted" and "Only duplicates"). We expect this feature to be extremely useful, as it is quite powerful; for example, by selecting all sources and the "Only duplicates" option, it allows to "find" all items marked as duplicates, which is a functionality that might be very useful, on occasion.
New Feature: search by score
In EPPI-Reviewer 4, when a search comes with "scores" (happens for classifier-driven searches, mostly), from the window that "visualises" the scores distribution, it is possible to create a new search with "score threshold(s)". This release includes the same functionality in the corresponding panel for EPPI-Reviewer Web.
Enhancement: sorting comparisons
Big reviews can contain hundreds of coding assignments and comparisons. While EPPI-Reviewer Web allowed to sort assignments, comparisons were still shown in a fixed order, based on Date. Now it is possible to sort the comparisons list by reviewer, coding tool, (optional) filter code and Date.
Enhancement: Search on Mag (EPPI-Reviewer Web)
The forms to pick year threshold(s) when searching for items in the Microsoft Academic Graph data are now improved (before, they showed a full "date" dialog, which was both fiddly and confusing).
Bugfix: importing from PubMed
In "import items\PubMed" it is possible to search PubMed directly from within EPPI-Reviewer 4 and Web. However, the Web version had some problems with its implementation. These were somewhat intricate, but showed up in practice in two ways: (1) when a PubMed search returned one result only, it was difficult to "convince" EPPI-Reviewer Web to actually import that result. Moreover, (2) when asking to either preview or import a subset of the search results, things could go wrong and result in importing/previewing a list of items that was shifted by a factor of 1 item. This problem is now resolved.
Bugfix: changing code-type while in the "Item Details" page
A previous update introduced a new bug in EPPI-Reviewer Web: if someone used the "edit code" mini dialog in "Item Details", changed the code "Type" and clicked "Update", this made the user interface (temporarily) "forget" all the coding applied to the current item. The problem was easily solved by browsing to the next item and back, but could generate spurious errors if/when people tried to add coding to the item (without noticing the coding had disappeared). This problem is now solved.