These technical notes are not for the faint-hearted! They provide a transparent record of how the searches were technically constructed. The driving objective behind the design of the Palliative Care PubMed Searches was to provide users with a range of clinical topics restricted to the palliative care context. The searches represent a combination of the best available evidence and expert opinion about how to identify literature relevant to palliative care on the publicly available PubMed database.
PubMed actually comprises two sets of data. Most records are indexed for the Medline dataset and have been assigned formal MeSH terms. The remaining are those indexed for PubMed and have NOT been assigned MeSH terms. A component of these PubMed records will eventually be indexed for Medline, yet this is not true for all records.
A MeSH-based search of PubMed will therefore only search for Medline records, and will exclude many records, including those published recently which may not yet have been assigned MeSH terms. To maximise retrieval a search must look for the relevant articles in both datasets. Thus it was necessary to create a topic search that searched both datasets. In Boolean terms, this may be logically represented as Part 1 OR Part 2, where Part 1 is designed to search for Medline records using MeSH terms, and Part 2 is designed to search for PubMed records using textwords.
Part 1 consists of Concepts, logically represented as Part 1 = (Concept 1 AND Concept 2), where Concept 1 is designed to search for the topic, typically using one or two MeSH terms and Concept 2 is the palliative filter. [1-2]
Part 2 consists of 3 Concepts, logically represented as Part 2 = (Concept 3 AND Concept 4) NOT Concept 5, where Concept 3 is designed to search for a topic, typically using one or two textwords. Concept 4 is the palliative filter. Concept 5 is the standard limit for searching only Medline. By choosing to exclude these records, Part 2 of the search is effectively limited to items that are not in Medline, that is, PubMed records.
Note that because the Palliative Care Search Filter originally used MeSH terms, it has been modified to search for each term by changing the terms to [textword] rather than [MeSH]. Palliative care[MeSH] and hospices[MeSH] became superfluous (as they are included in the broader truncated textword searches of palliat* and hospice*) so were removed. The remaining terms were retained as textwords with one exception and one variation. Death[majr:noexp] was potentially too broad (reflected in the original choice of delimiters), so it was deleted to reduce retrieval of irrelevant articles. Advance care planning[MeSH] was truncated to advance care plan* to improve retrieval for this term.
About the Palliative Care Search Filter
The modification to the Palliative Care Search Filter (Concept 4) is based only on the key concepts originally included in the Palliative Care Search Filter. Changes were in accord with what we thought would provide the best balance between sensitivity, specificity and precision. There are a range of other textwords that could be used to further increase sensitivity, such as end of life[textword], terminal illness[textword], life sustaining[textword] etc. For further suggestions, refer to the Palliative Care Search Strategy included in the Collaborative Review Group details on the Cochrane Library for the Cochrane Pain, Palliative and Supportive Care Group.
The Palliative Care Search Filter was designed and tested using the OVID version of Medline. It was subsequently translated into the correct syntax for PubMed. CareSearch uses this translated version, combined with a list of the palliative specialty journals, for reasons provided in subsequent research. 
In order for a journal to be included it had to be in English, with a specialty palliative care focus, and indexed on PubMed. A list of key journals that indexed palliative care literature identified through a national reference group of palliative care professionals as part of a previous project was used as the basis for a three part strategy designed to identify appropriate journals. First, all palliative care and supportive care specific journals were extracted from this list. Second, the List of Journals indexed on PubMed Central and Medline was searched via the US National Library of Medicine, using the terms palliat*, hospice*, supportive, death, dying, symptom*, quality of life, end of life, grief, griev*, bereav*, morbidity, mortality, pain and terminal. Third, these search terms were also used in Ulrich’s Periodical Database. Journal titles were combined from all three sources into a list which was then independently reviewed by a medical consultant, and two researchers, who agreed on the subset which met the selection criteria.
Titles include: American Journal of Hospice & Palliative Care; American Journal of Hospice Care; BMC Palliative Care; Death Education; Death Studies; Hospice Journal; International Journal of Palliative Nursing; Journal of Pain & Palliative Care Pharmacotherapy; Journal of Pain & Symptom Management; Journal of Palliative Care; Journal of Palliative Medicine; Journal of Social Work in End of Life & Palliative Care; Journal of Supportive Oncology; Omega; Palliative & Supportive Care; Palliative Medicine; Supportive Care in Cancer.
About additional concepts
Palliative Care PubMed Searches include further concepts, as chosen by the users. This could be a restriction to systematic reviews or randomised controlled trials, or to a specific publication date range eg, last 3 months, or to freely available full text links. CareSearch uses the syntax provided by PubMed for each limit in its exact form.
- For free full text, this means free full text[sb]
- For systematic reviews, this means systematic[sb]
- For last 3 months, this means “last 3 months”[dp]
- For randomised controlled trials, this means the sensitive filter for therapy (refer to the home page for PubMed Clinical Queries for more information).
The clinical and practice topics relate to pages developed for inclusion in the website. They were selected following feedback from the National Advisory Group and a review of national guidelines and health and social initiatives.
About variant searches offered on CareSearch
Five standard versions of each Palliative Care PubMed Search are provided in CareSearch:
- Strongest evidence (limited to systematic reviews or randomised controlled trials)*
- Everything (limited to last 3 months)
*An option restricting this search to free full text articles only is provided.
About language restrictions
Searches are restricted to English language. Language is a core field in the PubMed database, so both Medline and PubMed records will include this data.
Searches are not restricted to humans. The term ‘human’ is a MeSH term, so any strategy incorporating its use would exclude all PubMed record without MeSH indexing. It could technically have been included in Part 1, yet we decided not to do so for the sake of consistency across the overall topic search. Consequentially, searches may retrieve references to articles relating to animals, but we anticipate this will be a small and inconsequential number.
About sensitivity and specificity
The selection of MeSH terms used in Concept 1, and textwords used in Concept 3, are based on our expert opinion. In general we focused on higher specificity at a probable expense to sensitivity. This was because we didn’t want people to optimistically select a search topic only to find a large number or irrelevant articles. However this also means that searches are not designed to be comprehensive, so some relevant references may be missed.
For the exact syntax used in the Palliative Care PubMed Searches, see Palliative Care PubMed Search Syntax (75kb pdf). For the exact syntax used by the palliative care filter, see Palliative Care Search Filter Syntax (68kb pdf).