Gold Standard Search
Gold Standard Search
Thank you to everyone who mailed in search strategies in response to the "Gold Standard Search" article in the January-February 1997 issue of the Technical Bulletin. This feature offers searchers a chance to match wits with NLM staff on a search strategy formulation. The Gold Standard search strategies are developed here at NLM. The search request published in the January-February 1997 issue was:
Does zinc work in treating the common cold?
There were two winners for this search and one honorable mention. Sally A. Kilby, RN, M.S., of California Hospital Medical Center in Los Angeles, CA and Caroline Hopkinson of Lane Library in Savannah, GA both submitted strategies good enough to be winners and as you will see, they were not the same. Cheryl Banick of Warwick, Rhode Island receives an honorable mention for her careful research of the subject.
Choosing a File
The first step in this Gold Standard Search (see Figure 1) was to choose the best database for the subject. For this search, it was necessary to select the MEDLINE files. One person chose HealthSTAR in addition to MEDLINE, but because the subject of this search is purely clinical, that is not a good choice for this search. Only Ms. Kilby remembered to go into PREMEDLINE and OLDMEDLINE to search for the most recent citation and the oldest electronic citation on the subject.
Analysis of the Search Strategies
The selection of the appropriate MeSH headings was important. In this search the two obvious headings are ZINC and COMMON COLD. Most of the submitted searches included these headings; however, many searches used common (tw) and cold (tw) and one person used common@cold. When there is a MeSH heading that exactly covers the concept you are seeking, it is not a good idea to do this. Using the proximity feature adjacency is a particularly costly and inefficient way to retrieve data for which there is a MeSH Heading. Searching for common cold (mh) retrieved 162 citations almost instantly and cost $.06; searching common adj cold retrieved 146 citations after a pause of a few seconds and cost $.39. You are paying more money to make the computer work harder and retrieve less if you use adjacency inappropriately.
Ms. Kilby searched directly in ELHILL using the MeSH headings ZINC and COMMON COLD. She saved the strategy she ran in MEDLINE and used this strategy in all the Backfiles. She also searched in PREMEDLINE combining zinc and common and cold as Text Words and she went back to OLDMED where she searched on the above three words as Keywords (KW). There were no citations in these files, but she gets a gold star for remembering to look there. See Figure 2 on page 14.
Ms. Hopkinson used Grateful Med Version 6.7 for DOS and she went beyond COMMON COLD, to include the following MeSH Headings: COUGH, SNEEZING, HOARSENESS, RHINITIS, and RHINOVIRUS. As she noted these terms did not retrieve any additional hits when paired with ZINC and the other headings on that subject line; however, the effort to include additional relevant MeSH headings is noteworthy. See Figure 3 on page 15.
Ms. Banick, the honorable mention, ran her search on a CD-ROM version of MEDLINE. She did a particularly thorough job of researching which terms to include, using the scope note for COMMON COLD from MeSH, as well as consulting Taber's Cyclopedic Medical Dictionary and Dorland's Illustrated Medical Dictionary. Her investigations led her to include RHINOVIRUS in her search strategy, but to exclude other zinc compounds because she concluded from her readings that only zinc, and not zinc compounds, were used for human consumption. However, at least one relevant citation (97178330) is indexed with the MeSH Heading ZINC SULFATE. So she would have missed this citation, although there was a reasonable thought process behind the decision.
The Gold Standard search was run with MeSH Headings as a MultiFile search in MEDLINE and the Backfiles with a separate Text Word strategy for PREMEDLINE and OLDMEDLINE. Text Words were used in PREMEDLINE because MeSH headings have not yet been assigned to citations in PREMEDLINE. Text Words were also used to search OLDMEDLINE, which does not have up to date MeSH Headings and uses Keywords (the MeSH Headings that were to cover the subjects. It is possible to use Keywords (KW) to look for the MeSH Headings that were in use then or to search on Keyword Fragments (KF) to look for individual words in the old MeSH Headings. Searching with Text Words will look for the words in the Keyword field (in the same way that searching qualified by KF does); however, Text Words will also be searched for in the Title and Abstract fields. Text Word searching is more comprehensive than searching by Keyword Fragments and it doesn't require that you check on the MeSH Headings used prior to 1966 (which you should do before searching on Keywords.)
Search Strategy Suggestions
The most common error was the use of subheadings in a search that had a relatively small retrieval. If you find fewer than 10 citations in a file, it is generally unnecessary to further restrict the retrieval by including subheadings.
The Next Gold Standard Search
The search request is:
A physician is trying to decide whether or not to recommend yearly mammographies to her 40-49 year old patients and wants to get some information to help her make this decision.
Results will be published in the September-October 1997 Technical Bulletin. Entries must be received by August 22, 1997 to be considered.
Mail us a copy of your search strategy and include a printout of citations. Please note that in the interest of saving time, money, and paper we are no longer asking for abstracts. If your strategy is selected for publication you could win one hour of free search time for non-Web users or an NLM T-shirt and mug. In cases of multiple submissions of identical winning strategies, the prize winner will be either the person who is a first time winner or the first winning strategy received.
Search Hints for Direct Searchers
Select your database carefully. Limit your retrieval to English language citations.
There are no exact answers or absolutely correct strategies, only best approximations, just like real life. Good luck!
Please send a copy of your search strategy with the retrieval, which database was searched, and whether you searched using command language, Grateful Med, Internet Grateful Med, or PubMed to:
SS 1 /C? USER: file premedline PROG: YOU ARE NOW CONNECTED TO THE PREMEDLINE FILE. NOTICE: THESE PRELIMINARY, UNINDEXED CITATIONS HAVE NOT GONE THROUGH NLM'S QUALITY CONTROL PROCESS. SS 1 /C? USER: (tw) zinc and common and cold * Searching by Text Word because MeSH headings have not been assigned to PREMEDLINE citations. PROG: *NONE- SS 2 /C? USER: file oldmed PROG: YOU ARE NOW CONNECTED TO THE OLDMEDLINE (PRE-1966 CUMULATED INDEX MEDICUS) FILE. SS 1 /C? USER: (tw) common and cold and zinc * Search by Text Word, there are no updated MeSH Headings in OLDMEDLINE. PROG: *NONE- SS 2 /C? USER: mfs mode=auto * Use MultiFile Searching to search all years of MEDLINE (1966 - present). PROG: SEARCH TO BE RUN ALREADY EXISTS. DO YOU WISH TO REPLACE IT? (Y/N) USER: y PROG: MULTIFILE SESSION ABOUT TO BEGIN MODE: AUTO SEARCHNAME: &-MFS-& (TO BE INPUT) FILES TO BE SEARCHED: ** MEDLINE (1994 - 97) ** BACK90 (1990 - 93) ** BACK85 (1985 - 89) ** BACK80 (1980 - 84) ** BACK75 (1975 - 79) ** BACK66 (1966 - 74) OK? (Y/N/C/LIST) USER: y PROG: ENTER SEARCH- YOU ARE NOW CONNECTED TO THE MEDLINE (1994 - 97) FILE. *STS* SS 1 /C? USER: zinc PROG: SS (1) PSTG (2639) *STS* SS 2 /C? USER: common cold or rhinovirus or rhinitis PROG: SS (2) PSTG (639) *STS* SS 3 /C? USER: 1 and 2 PROG: SS (3) PSTG (10) *STS* SS 4 /C? USER: prt 3 ti * Print a few titles to check the quality of the retrieval. PROG: * The three letters shown here are unique despite having identical titles. 1 TI - Zinc gluconate lozenges for treating the common cold [letter] 2 TI - Zinc gluconate lozenges for treating the common cold [letter] 3 TI - Zinc gluconate lozenges for treating the common cold [letter] *STS* SS 4 /C? USER: 3 and not for (la) PROG: SS (4) PSTG (10) *STS* SS 5 /C? USER: finished * "Finished" ends the search in MEDLINE and cues the computer to run the strategy in the Backfiles. PROG: SEARCH SET FROM SS 1 FORWARD SAVED AS '&-MFS-&'. YOU ARE NOW CONNECTED TO THE BACK90 (1990 - 93) FILE. SS (2) PSTG (3) YOU ARE NOW CONNECTED TO THE BACK85 (1985 - 89) FILE. SS (3) PSTG (8) YOU ARE NOW CONNECTED TO THE BACK80 (1980 - 84) FILE. SS (4) PSTG (1) YOU ARE NOW CONNECTED TO THE BACK75 (1975 - 79) FILE. SS (5) PSTG (1) YOU ARE NOW CONNECTED TO THE BACK66 (1966 - 74) FILE. TIME OVFLW: CONT? (Y/N) USER: y PROG: SS (6) PSTG (4) READY TO PRINT FROM SEARCH 1, CONTAINING 10 ITEM(S) FROM MEDLINE. ENTER SPECIFICATIONS FOR PRINTING (E.G., BR, SD, FU, DL), OR 'SAME' (DO NOT USE THE WORDS 'PRINT' OR 'PRT'): USER: br * Select BR print option to browse the titles. PROG: 1 (MEDLINE) TI - Zinc gluconate lozenges for treating the common cold [letter] 2 (MEDLINE) TI - Zinc gluconate lozenges for treating the common cold [letter] 3 (MEDLINE) TI - Zinc gluconate lozenges for treating the common cold [letter] 4 (MEDLINE) TI - Zinc lozenges to treat colds [letter; comment] 5 (MEDLINE) TI - Zinc lozenges as cure for common colds. 6 (MEDLINE) TI - Zinc for treating the common cold: review of all clinical trials since 1984. 7 (MEDLINE) TI - Zinc gluconate lozenges for treating the common cold. A randomized, double-blind, placebo-controlled study [see comments] 8 (MEDLINE) TI - Zinc: the biology and therapeutics of an ion [editorial; comment] 9 (MEDLINE) TI - How does zinc modify the common cold? Clinical observations and implications regarding mechanisms of action. 10 (MEDLINE) TI - The 2A proteinase of human rhinovirus is a zinc containing enzyme. READY TO PRINT FROM SEARCH 2, CONTAINING 3 ITEM(S) FROM BACK90. REPRINT PRIOR SEARCH RESULTS? (Y/N/C/SKIP PLUS OPTIONAL WORD(S) 'PRINTSPECS' AND/OR 'RECORDS' TO REQUEST PROMPTS FOR THESE CHANGES): USER: n * Type N to skip reprinting and see citations in the next file. PROG: 1 (BACK90) TI - Zinc lozenges for treatment of common colds. 2 (BACK90) TI - Zinc gluconate and the common cold: a controlled clinical study. 3 (BACK90) TI - Vitamins, minerals, and the common cold. READY TO PRINT FROM SEARCH 3, CONTAINING 8 ITEM(S) FROM BACK85. REPRINT PRIOR SEARCH RESULTS? (Y/N/C/SKIP PLUS OPTIONAL WORD(S) 'PRINTSPECS' AND/OR 'RECORDS' TO REQUEST PROMPTS FOR THESE CHANGES): USER: finished * Typing "finished" will cancel the remainder of the printout. PROG: SUPERPRINT COMMAND HAS BEEN CANCELLED.
Editor's Note: Sally also searched in PREMEDLINE and OLDMEDLINE. The PREMEDLINE search is identical to the Gold Standard Search and is not being reprinted here.
SS 1 /C? USER: file oldmed PROG: YOU ARE NOW CONNECTED TO THE OLDMEDLINE (PRE-1966 CUMULATED INDEX MEDICUS) FILE. SS 1 /C? USER: (kw) common cold and zinc PROG: *NONE- SS 1 /C? USER: file med YOU ARE NOW CONNECTED TO THE MEDLINE (1994 - 97) FILE. SS 1 /C? USER: zinc PROG: SS (1) PSTG (2639) SS 2 /C? USER: common cold PROG: SS (2) PSTG (162) SS 3 /C? USER: 1 and 2 and not for (la) PROG: SS (3) PSTG (9) * At this point you could give a PRINT command to see the retrieved citations. SS 4 /C? USER: save PROG: SEARCHNAME? USER: zinccold PROG: SEARCH SET FROM SS 1 FORWARD SAVED AS 'ZINCCOLD'. SS 4 /C? USER: file med90 PROG: YOU ARE NOW CONNECTED TO THE BACK90 (1990 - 93) FILE. SS 1 /C? USER: zinccold (sn) PROG: SS (1) PSTG (3) SS 2 /C? USER: file med85 PROG: YOU ARE NOW CONNECTED TO THE BACK85 (1985 - 89) FILE. SS 1 /C? USER: zinccold (sn) PROG: SS (1) PSTG (6) SS 2 /C? USER: file med80 PROG: YOU ARE NOW CONNECTED TO THE BACK80 (1980 - 84) FILE. SS 1 /C? USER: zinccold (sn) PROG: SS (1) PSTG (1) SS 2 /C? USER: file med75 PROG: YOU ARE NOW CONNECTED TO THE BACK75 (1975 - 79) FILE. SS 1 /C? USER: zinccold (sn) PROG: *NONE- SS 1 /C? USER: File med66 PROG: YOU ARE NOW CONNECTED TO THE BACK66 (1966 - 74) FILE. SS 1 /C? USER: zinccold (sn) PROG: SS (1) PSTG (1)
Carol Hopkinson performed her search with Grateful Med for DOS; however, we are showing the search as it would display in our new Windows upgrade.