How would do manage over 10,000 citations in synch across users in different time zones?

Analyzing and Reusing Data
1

I have been an EndNote freak for years and recently, Zotero. In conducting a systematic review of a sort, what citation management tools or best practices do you follow? What has worked best for you in keeping tabs along the process? Would you recommend LLMs for titles and abstracts screening, data extraction? How do LLMs compare with the human intervention in making judgement on articles objectively, given the eligibility criteria? HELP ME!!!

2 Likes
2

i’m a sucker for mendeley for reference software… i appreciate that you can add papers easily and use it on different computers with your log in and all. zotero has the google drive plug in though, so that’s a big appeal but i haven’t been able to convince myself to make the switch because i’m a weirdo who struggles with change…

for systematic reviews, i’ve used rayyan and now i’m using covidence. if anyone in your review team has the institution paying for covidence, it’s pretty nice so far. it’s not free for most features though, but you only need one person to have the paid version, and then you can invite others without having them pay. although it absolutely takes a lot of time, i find the title/abstract screening phase to be a fabulous excuse for you to read so many articles actually. so i like the human aspect of it, haven’t attempted to do it with LLM at all and haven’t seen pubs using LLM for systematic review yet, but that may be a me problem

2 Likes
3

I generally use DistillerSR but it’s really expensive. Covidence is probably the best alternative for managing multiple reviewers. I’ve seen demos of other lit review management tools (Moara, Sysrev, etc.) that incorporate LLMs but haven’t used them.

As far as where LLMs are for screening and extraction, I would say they are pretty good at screening if you give them the right prompts and good inclusion/exclusion criteria. You’ll need to do some calibration on a sample but once you’re satisfied with that the results can be as good as human-only. Data extraction is a lot more iffy at this point, but it depends on what elements you’re looking for. See https://effectivehealthcare.ahrq.gov/products/machine-learning-tools/white-paper

4 Likes
4

Thanks for your input, @ecebayram. I like that I’ll get to read 10 papers in an hour :grinning_face: .

1 Like
5

Now that I have Covidence mentioned twice, I can only dive into it. Thanks for sharing your thoughts!

1 Like