Searching PubMed for all the Evidence: A Review for PHPR 622 (cont)

If you haven't already completed part I of this tutorial, click here to go to part I.

 

If you've finished part I, you've already created:

  • a MeSH search for trials concerning use of eculizumab to treat paroxysmal nocturnal hemoglobinuria (PNH) with thrombosis
  • a core keyword search concerning use of eculizumab to treat PNH.

You're now ready to test your keyword search in PubMed.

Keyword -- 9. Run the Search in PubMed

1 of 2

Now that you've constructed your keyword search you need to return to PubMed.

  • Click here to open PubMed in a new window.
  • Paste your search strategy into the search box.
  • Hit the "Search" button.
  • Check to be sure you removed the filters used for the MeSH search. If the filters are still present, click on the "clear all" link now (see arrow in figure below).


A screenshot of the "Clear all" link in PubMed.

Keyword -- 9. Run the Search in PubMed

2 of 2
  • Review highlighted messages above the result list

If you used truncation with a trunk shorter than 4 letters, you will be told that you need to lengthen the word trunk.

Screenshot of a "The following term was not found in PubMed" message.

If single-word or hyphenated terms are not found in any PubMed record, a message to that effect will appear (see arrow in figure).

Check the terms listed for mis-spelled search terms. Correct any mis-spellings in your search statement. 

Do not remove single-word or hyphenated terms from your search just because they're "not found."  You may need to update your search in the future, and these terms may be present in a record added to PubMed tomorrow.

PubMed will say that the [all] was ignored, but it stops autocorrection none-the-less. Do not remove it.


A message about "quoted phrases" that were "not found" is not produced by the tutorial search, but may be present when you work on your project searches.

This does not mean that there were no records in PubMed containing the listed quoted phrases.  It means that the listed quoted phrases were not in the list of phrases that the powers-that-be have decided PubMed should search as phrases. 

If a space-containing drug name is in the list of "not found" "quoted phrases,"  remove the drug name from your search and keep a record of the removed term/s.  The way PubMed handles "not found" phrases causes problems when the "not found" phrase is a drug name.

If a disease name or other non-drug concept that contains a space is not found, it's usually safe to leave the phrase in your search. 

Sometimes a "Did you mean" message is present.  Pubmed often offers you an errant, autocorrected version of your search. DO NOT click on this altered version of your search.

Keyword -- 10. Check the "Search Details" Page

1 of 2

Even if you see no warnings, it's a good idea to look at the "Search Details."

  • Click the "Advanced" link under the search box.
A screenshot shows the position of the "Advanced" link.
  • Scroll down to the "History and Search Details" table.
  • Click on the arrow in the "Details" column.

A screenshot shows the position of the "Details" column in the "History and Search Details" table.

  • Scroll down to look at the "Translations"
A screenshot of the "Translations"

The "translations" of terms in this search are all true to the intended meaning of the entered keywords. There is no reason to make any alterations in the search strategy because of these translations.

  • If inappropriate translations are a problem in your future searches, add title/abstract tags -- [tiab] -- after the inappropriately translated terms. This stops translation of the tagged terms. Truncating a term or enclosing the term in quotes will also stop translation.

Keyword -- 10. Check the "Search Details" Page

2 of 2If you're using this tutorial to guide the completion of your group project search or another non-tutorial-related search, you may have made some revisions in your search based on the messages above the search results or based on the "Translations" list.
  • Scroll up to the "Query Box"
  • Paste your revised search into the box.
  • Click the "Search" button.
  • Check the messages above the results list and the "Details" on the "Advanced" page again.

Keyword -- 11. Three Versions of Your Search

You will be required to run three versions of the keyword search that you  create for your group project:
  • a keyword search for previously written systematic reviews,
  • a keyword search for un-indexed records (the records that cannot be retrieved by a MeSH search) and
  • a search for indexed studies.  You can apply publication type limits to this search.  If your project topic concerns therapeutic efficacy as the tutorial search does, randomized controlled trials would be the best evidence.  This search can be limited to dates that are not covered by the MeSH search.

This tutorial will give you a chance to practice producing these three types of keyword searches using the "eculizumab treatment of PNH" search.

Keyword -- 11 a. "systematic rev.+ meta-analyses + guidelines" Subset

1 of 2

What is retrieved?

The first required variant of the keyword search is a search for indexed and unindexed:

  • guidelines,
  • consensus statements,
  • systematic reviews, and
  • meta-analyses.

Why?

Authors of new systematic reviews are expected to be aware of the recommendations made in these previous publications.  Drs. Yee and Cochran like to ask how group findings compare to recent practice guidelines.

How?

  • Click on UNMC's custom "systematic rev.+ meta-analyses + guidelines" filter
A screenshot of the filters on the left-hand side of the UNMC PubMed page.

Keyword -- 11 a. "systematic rev.+ meta-analyses + guidelines" Subset

2 of 2Disadvantages of the "systematic rev.+ meta-analyses + guidelines"  filter:
  • This filter is only available when you reach PubMed through a link on UNMC's website.
  • This filter will be removed every time you do something to your search that makes the search reload.
  • This filter will not show up in your search history.  When you work on the search strategy for your group project, you will have to type --

"systematic rev.+ meta-analyses + guidelines" filter

-- under the keyword search strategy (with any other filters applied) in the section of the search strategy template devoted to systematic reviews. 

  • This filter is not available through the "Article types" area on the right.  The "systematic reviews" filter on the left now retrieves systematic reviews only.  It, like the "systematic rev.+ meta-analyses + guidelines"  filter and unlike the other article type filters, will retrieve indexed and unindexed records. 
 
How do the "systematic rev.+ meta-analyses + guidelines"  and "systematic review" filters retrieve indexed and unindexed records?

Both are actually  search strategies that are AND'd (invisibly) to your search.  They  include OR'd together keywords, journal titles, publication type tags and MeSH headings that are characteristically found in records for the publication types of interest.  Because keywords and journal titles are included in the filters, they can retrieve unindexed and indexed records.

 

Keyword -- 11 a. Systematic Subset Troubleshooting

1 of 2

If your search is retrieving no results:

  • Check the left hand column of the results page. If you previously selected any limit other than English, click on the limit. Wait for the page to refresh and the check mark to disappear.
  • Check your search statement. Focus on looking for missing OR's and errant ANDs.
  • Remember that you can ask a librarian for help.
  • If you are working on a non-tutorial search topic, remember that not all topics have guidelines or systematic reviews.

Keyword -- 11 a. Systematic Subset Troubleshooting

2 of 2

If your search is retrieving far more irrelevant than relevant results:

  • Check the messages above the results list for multi-word (space-containing) terms that are "not found" by PubMed. Consider removing the "not found," multi-word search terms from your search statement. Save a record of any removed terms.Check your search statement for unmatched parentheses and missing sets of parentheses.
  • Check your search statement for errant ORs (ORs where ANDs should be)
  • Scan the first page of results for bold-faced search terms that almost always have a meaning other than your desired meaning. If the terms are not absolutely essential to your search, you can remove them from the search statement. Save a record of any removed terms.
  • Click the "Display Options" button and use the "Format" drop-down menu to select the "Abstract" view. Look for author's names and institutional affiliations that contain your search terms. In your search statement, add [tiab] after any search terms that are found in the author and affiliations fields. This will restrict these terms to title/abstract searching.
  • Remember almost all comprehensive keyword searches retrieve quite a few irrelevant results.
  • Remember that you can ask a librarian for help.

As of 8/2021, the "paroxysmal hemoglobinuria and eculizumab," systematic review search retrieves only 15 results (14 English-language results). Two of these, a 2021 J Pediatr Hematol Oncol article and a 2014 Cochrane Database of Systematic Reviews article would be very useful to a student writing his or her own systematic review on this topic.

Keyword -- 11 b. Unindexe Records

What is retrieved?

The second version of your keyword search that must be run retrieves unindexed records.  Including:

  • PREMEDLINE records.  These are records waiting to be indexed.
  • Non-MEDLINE, PubMed Central records.   These are records for full-text articles from journals not indexed by MEDLINE.  They are represented in PubMed because the full-text versions have been deposited in PubMed Central.

Why?  The most up-to-date articles on your topic may be represented by records that haven't been indexed yet. These records can't be retrieved by a MeSH search.  You wouldn't want to miss the most up-to-date evidence.

 

How?

  • Place your cursor at the end of your search statement, add a space, and then add --

NOT MEDLINE[sb]
-- to your search statement (as shown below).

  • If yo wish, you can also add -- 
AND random*
-- to the end of your search. 
 
You MUST NOT add an actual "Randomized Controlled Trials" "Article Type" limit because searches with that filter only retrieve indexed records that contain the indexer-applied "randomized controlled trials" tag.  The closest you can come to adding an RCT filter to a search for unindexed records is adding the random* keyword.   Why does this work?  It's hard work to randomize a trial.  Therefore, all (or very nearly all) authors of articles describing randomized controlled trials will include a form of the word random (random, randomized, etc.)  in either the title or abstract.    Not all results retrieved will be randomized controlled trials some will just discuss RCTs.
 
What if your group has to include non-randomized studies in your systematic review?  Obviously, you will not be able to add -- AND random* -- to your search.  You will just have to browse through all the titles and abstracts of the records retrieved by your search for unindexed records.  Look for records that might meet your review's inclusion criteria.  You only have to read the full-text of those that look like they might meet your criteria.
  • Hit the "Search" button.

Keyword -- 11 b. Troubleshooting Unindexed

1 of 3If your search is retrieving no results:
  • Check the left hand column of the results page. If you have selected any limit other than English, click on the limit. Wait for the page to refresh and the check mark to disappear.

(Publication date limits can be used when searching for unindexed records, but there are few good reasons to use publication date limits. Think twice before applying a pub date limit.)

  • Check your search statement. Focus on looking for missing OR's and errant ANDs.
  • Remember that you can ask a librarian for help.
  • If you are working on a non-tutorial search topic, remember that not all topical searches will retrieve unindexed records.

Keyword -- 11 b. Troubleshooting Unindexed

2 of 3If your search is retrieving far more irrelevant than relevant results:
  • Check the messages above the results list for multi-word (space-containing) terms that are "not found" by PubMed. Consider removing the "not found," multi-word search terms from your search statement. Save a record of any removed terms.Check your search statement for unmatched parentheses and missing sets of parentheses.
  • Check your search statement for errant ORs (ORs where ANDs should be)
  • Scan the first page of results for bold-faced search terms that almost always have a meaning other than your desired meaning. If the terms are not absolutely essential to your search, you can remove them from the search statement. Save a record of any removed terms.
  • Click the "Display Options" button and use the "Format" drop-down menu to select the "Abstract" view. Look for author's names and institutional affiliations that contain your search terms. In your search statement, add [tiab] after any search terms that are found in the author and affiliations fields. This will restrict these terms to title/abstract searching.
  • Remember almost all comprehensive keyword searches retrieve quite a few irrelevant results.
  • Remember that you can ask a librarian for help.

Keyword -- 11 b. Troubleshooting Unindexed

3 of 3As of 8/2021, the search for unindexed records concerning paroxysmal hemoglobinuria and eculizumab retrieved:

100 results before addition of the -- AND random*. 

4 results after addition of the  -- AND random*

 

Keyword -- 11 c. Randomized Controlled Trials

What is retrieved?

  • The third required keyword search retrieves MEDLINE (indexed) records for randomized controlled trials or other type of study/ies that are the focus of your systematic review.

Why? 

  • The RCT(other study type)-limited keyword search will cover dates before the introduction of subject headings used in your MeSH search.
  • The search will allow you to check the adequacy of your MeSH search.

How:

  • Remove the -- NOT MEDLINE[sb] -- from the end of the search statement .
  • Hit the "Search" button to get full results before application of limits.
  • If the "English" filter hasn't already been chosen, click on it.  You may have to use the "Show additional filters" link at the bottom of the left-hand column to select and show the language limits before you can select the "English" limit.
  •  Click the "Randomized Controlled Trials" and the "Humans" filters.

Click on the filters one at a time. Each time wait for the page to refresh and limited results to appear before clicking on the next filter.

Keyword -- 11 c. Randomized Controlled Trials Troubleshooting

1 of 3

If your search is retrieving no results:

  • Check your search statement to make sure you have removed the -- NOT MEDLINE[sb] -- from the end of the search statement. 
  • Check your search statement. Focus on looking for missing OR's and errant ANDs.
  • Remember that you can ask a librarian for help.

Keyword -- 11 c. Randomized Controlled Trials Troubleshooting

2 of 3If your search is retrieving far more irrelevant than relevant results:
  • Check the messages above the results list for multi-word (space-containing) terms that are "not found" by PubMed. Consider removing the "not found," multi-word search terms from your search statement. Save a record of any removed terms.Check your search statement for unmatched parentheses and missing sets of parentheses.
  • Check your search statement for errant ORs (ORs where ANDs should be)
  • Scan the first page of results for bold-faced search terms that almost always have a meaning other than your desired meaning. If the terms are not absolutely essential to your search, you can remove them from the search statement. Save a record of any removed terms.
  • Click the "Display Options" button and use the "Format" drop-down menu to select the "Abstract" view. Look for author's names and institutional affiliations that contain your search terms. In your search statement, add [tiab] after any search terms that are found in the author and affiliations fields. This will restrict these terms to title/abstract searching.
  • Remember almost all comprehensive keyword searches retrieve quite a few irrelevant results.
  • Remember that you can ask a librarian for help.

Keyword -- 11 c. Randomized Controlled Trials Troubleshooting

3 of 3

As of 8/2021, the keyword search for RCTs focused on eculizumab use in paroxysmal nocturnal hemoglobinuria retrieves the 13 results.  1 more record than was retrieved  by the MeSH search.

It's a good idea to compare the results of the MeSH and keyword search by looking at:

keyword results not retrieved by the MeSH search

and

MeSH search results not retrieved by the keyword search.

 

Keyword -- 12. Add search strategies to the "search strategy figure" template

Click on the "Advanced" link under the PubMed search box to see the search history table.

If you were working on your group project, you would need to save the search strategies for the three, keyword search variants in the "search strategy figure" template that already contains your MeSH search strategy.

Click here to see the keyword search history how the specific searches are added to the "search strategy figure" template.

Flow Chart

In the past, Drs. Cochran and Yee have asked that students in past classes include a record retrieval and review flow chart in their written projects and oral presentations. They will probably ask that you the same. Click here to see an example of what the flow chart might look like for this search.

The general outline for the flow chart is available for download at http://www.prisma-statement.org/documents/PRISMA_2020_flow_diagram_new_SRs_v1.docx.

Assignment PHPR 562

Please, return to Canvas and take the short test to verify your completion of this tutorial.

You will have a chance to submit your comments on this tutorial on the following page.  We value your comments.

 



Certificate

Please enter your name and email address to retrieve a copy of your completed quiz.

You can enter multiple email addresses separated by commas. If you are doing this for a class, you may need to enter your instructor's email address also.

Feedback

What did you think of this tutorial?