On 2014 Dec 10, Kath Wright commented:

Other search filters are available from the InterTASC Information Specialists' Sub-Group Search Filter Resource at https://sites.google.com/a/york.ac.uk/issg-search-filters-resource/home

On 2013 Dec 27, Wichor Bramer commented:

To be a useful addition to a search strategy, sensitive filters should not be too unspecific, and specific filters not too insensitive. Developing a sensitive filter of 100% is no problem, if specificity is not considered, very general filter will find all articles, whether they are positives or negatives, but does not add value to a search, because it does not limit the number needed to read (NNR). And a filter with 100% specificity is easy to create, by taking a very rare phrase that only occurs in one or two positives, but then the sensitivity is almost zero, and the filter is useless.

Simon, Hausner, et al. developed sensitive filters for nurse staffing with a sensitivity of almost 100%, but with a precision of 0.3% the NNR is almost 300, and even for their precise strategy the NNR was 3, with a sensitivity of less than 50%. The authors conclude themselves that 'nurse staffing studies are difficult to identify', but that is true for almost every thematic issue. The low values their filters renders them useless in practice.

The fact that the most precise filter contains the MeSH term Outcome and Process Assessment (Health Care) is related to the fact that this is part of the search strategy of one of the systematic reviews (Kane RL, 2007) that was used to gather the positives, a review of which the presented search strategy is very difficult to read and not repeatable. http://www.ncbi.nlm.nih.gov/books/NBK38310/

On 2013 Dec 27, Wichor Bramer commented:

To be a useful addition to a search strategy, sensitive filters should not be too unspecific, and specific filters not too insensitive. Developing a sensitive filter of 100% is no problem, if specificity is not considered, very general filter will find all articles, whether they are positives or negatives, but does not add value to a search, because it does not limit the number needed to read (NNR). And a filter with 100% specificity is easy to create, by taking a very rare phrase that only occurs in one or two positives, but then the sensitivity is almost zero, and the filter is useless.

Simon, Hausner, et al. developed sensitive filters for nurse staffing with a sensitivity of almost 100%, but with a precision of 0.3% the NNR is almost 300, and even for their precise strategy the NNR was 3, with a sensitivity of less than 50%. The authors conclude themselves that 'nurse staffing studies are difficult to identify', but that is true for almost every thematic issue. The low values their filters renders them useless in practice.

The fact that the most precise filter contains the MeSH term Outcome and Process Assessment (Health Care) is related to the fact that this is part of the search strategy of one of the systematic reviews (Kane RL, 2007) that was used to gather the positives, a review of which the presented search strategy is very difficult to read and not repeatable. http://www.ncbi.nlm.nih.gov/books/NBK38310/

On 2014 Dec 10, Kath Wright commented:

Other search filters are available from the InterTASC Information Specialists' Sub-Group Search Filter Resource at https://sites.google.com/a/york.ac.uk/issg-search-filters-resource/home