Ability to have n/N percentage in export#398

I would love for NK to automatically calculate n/N percentages for dichotomous variables (or opt for this to happen) so that these percentages show up when you export data. They currently show up in Synthesis, but when I download extracted data I don’t see an option to have these percentages anywhere.

For context, this came up because I’m currently working on a project with lots of dichotomous variables, so for now I’m just manually inputting lots of calculation functions in Excel so that I can later input these into all the tables for the manuscript.

Which export is this in reference to? “Custom Tables” or “Extracted Data”? The latter is intentionally left as “raw” data for an analyst/statistician to work with. For the former, I agree this would be a nice-to-have for custom tables!

Mildly related: https://nested-knowledge.nolt.io/382

Sure, this would be great to have in Custom Tables!

A question about Custom Tables, though––is there a way to make a custom table that actually includes extracted data? I played around with Export and also checked this wiki article (https://wiki.nested-knowledge.com/doku.php?id=wiki:autolit:extraction:tables), but as far as I can tell the only way to get extracted data is to use the Extracted Data tab. If Custom Tables had extracted data, this would speed up the process of making manuscript tables considerably!

Also, in the Extracted Data tab, it would be nice to filter to only Included studies rather than being forced to download all records that have at least one extracted data point. I’m currently working on a project that has been downsized a couple times so there are quite a few records with extracted data points that aren’t relevant to writing the manuscript anymore. When I download the Excel file, I have to go through and either filter out or manually delete all the Excluded studies.

is there a way to make a custom table that actually includes extracted data?

Yes, if you specify Table of: Study Arm, you will have Data Element export options. Since extracted metrics are scoped to study arms, more specifically than studies (the default).

it would be nice to filter to only Included studies rather than being forced to download all records that have at least one extracted data point.

I largely agree with this idea for conceptual clarity; I think we only added excluded studies to this table because of a one-off user request.

@Kevin.kallmes @Nicole_hardy @John Pederson Any objection to only exporting extracted (quantitative) data for included references?

Great, added this detail to the wiki! Thanks Karl

Yeah, I think that would be a good idea for a default option. I can think of couple times where users didn’t realize that excluded studies were in the downloaded dataset for extracted data and started doing analysis on both excluded/included studies (I did this once by accident). The vast majority of the time, users will be interested in only those included. There would be a workaround to this if they really do need to get the excluded data, but totally agree that only included would be the preferred default

^Same comment about exported RoB data. I would prefer the automatic setting to be only Included studies in the downloaded file. Thanks!

Yep, if we go that direction with one, it would be nice to go that way with both.

Counterpoint - sometimes a poor critical appraisal score is grounds for exclusion. Would users be upset to not have this data on hand?

I see it as a chicken-and-the-egg problem. I could imagine someone wanting the RoB information for a study they excluded after performing the RoB. But then again, if the study is ultimately excluded then it’s not going to be written up the manuscript, so it’s sort of irrelevant. That’s why I’d love to have options for these different data download scenarios

Issues with data collected for excluded studies usually arise when data is collected prior to rigorously assessing the quality of the overall study. As a workflow, its best to review studies in detail to ensure that they meet basic study criteria. Every once in a while, a mistake may be made and a study with collected data will be later excluded (the user should definitely change inclusion status to “Excluded” in this case). This is a different situation compared to when a study met baseline study criteria, moved on to RoB, and was excluded (from quantitative synthesis) due to very high risk bias. From a methodological perspective, it’s still best to document the full RoB results regardless of whether a study was excluded on the basis of risk of bias. While exclusion based on RoB is sometimes done in the literature (and some users will want this), it is not a recommended strategy, and the researcher should know if it makes sense to include a study based on their subject matter expertise. It’s also true that in a systematic review, the high risk of bias studies can be helpful to summarize some of the weaknesses/gaps in the literature, and to identify areas which should be further investigated (this is a good goal for any systematic review). However, I could see how this might be difficult to implement for qualitative and quantitative synthesis. This is getting off topic, but it would be ideal to have “include for qualitative synthesis” and “include for quantitative synthesis”

Thanks for your input guys! Based on this, we need to provide options for both included & all records in export (QLS/QNS is a separate issue I’ll leave for a different day).

No problem! Fyi, I just created a separate post about qualitative vs. quantitative synthesis inclusion status since this got pretty off topic from the original question

Closing, as I believe all needs here have been addressed (extracted data export has an included/all checkbox, we compute dichotomous %s in table builder). Let me know if we missed the mark or didn’t fully address!