On 2015 Jan 07, Wichor Bramer commented:

Volpato et al investigated in their article what the best method is to search phrases in the database PubMed. Does it matter whether parentheses, quotes or truncation are used? They could have saved themselves time and FAPESP money, because the answer to that can easily be found in NLM documentation, or by attending an advanced PubMed class. Parentheses are necessary only when combining OR's and AND's. Around phrases they do not change anything to the results. Quotes around phrases create an exact search for that phrase, while phrases without quotes will be split in parts combined with AND. Ending a phrase with an asterisk (truncation) will search for any phrase that starts with that phrase. In general most results will be found not using quotes or truncation (but this might contain much noise), followed by truncated phrases, and the least of quoted phrases. (It would be interesting to see the four queries in which quotes retrieved more hits). Therefor Volpato et al conclude in the text that no quotes should be used. However they state that quotes are recommended when the searcher wants to be exact (that is indeed the case, the search will be more exact) and they explicitely do not recommend using truncation, since it reduces the number of hits. But had they compared truncation to quoted phrases they would have found that truncated phrases generate more hits than untruncated phrases (cardiac event OR cardiac events), at least when applied well (in their example they should have used intracapsular partial tonsillectom*, to also find the intracapsular partial tonsillectomies). Whether or not the simple search or the search history tool should be used completely depends on the searcher. If it does not make a difference in the results, one can not draw a conclusion to that, and one should not recommend one or the other. A good systematic search, however, combines truncated phrases and mesh terms and is optimized to find as much relevant articles as possible.

On 2015 Jan 07, Wichor Bramer commented:

Volpato et al investigated in their article what the best method is to search phrases in the database PubMed. Does it matter whether parentheses, quotes or truncation are used? They could have saved themselves time and FAPESP money, because the answer to that can easily be found in NLM documentation, or by attending an advanced PubMed class. Parentheses are necessary only when combining OR's and AND's. Around phrases they do not change anything to the results. Quotes around phrases create an exact search for that phrase, while phrases without quotes will be split in parts combined with AND. Ending a phrase with an asterisk (truncation) will search for any phrase that starts with that phrase. In general most results will be found not using quotes or truncation (but this might contain much noise), followed by truncated phrases, and the least of quoted phrases. (It would be interesting to see the four queries in which quotes retrieved more hits). Therefor Volpato et al conclude in the text that no quotes should be used. However they state that quotes are recommended when the searcher wants to be exact (that is indeed the case, the search will be more exact) and they explicitely do not recommend using truncation, since it reduces the number of hits. But had they compared truncation to quoted phrases they would have found that truncated phrases generate more hits than untruncated phrases (cardiac event OR cardiac events), at least when applied well (in their example they should have used intracapsular partial tonsillectom*, to also find the intracapsular partial tonsillectomies). Whether or not the simple search or the search history tool should be used completely depends on the searcher. If it does not make a difference in the results, one can not draw a conclusion to that, and one should not recommend one or the other. A good systematic search, however, combines truncated phrases and mesh terms and is optimized to find as much relevant articles as possible.