Developing the search strategy is a process that takes time and care and expertise. The steps include:
The question framework provides the scaffolding for the search strategy. It defines the parameters of the studies being sought and provides the basis of the terms needed and the structure of the search string.
Early on in the search process a group of articles that represent the evidence that the search strategy will need to find should be identified. These articles are key to developing and testing the search strategy. They are equally important at the beginning of the search process and in the testing phase. At the start they are key sources to be mined for terms for building the search strategy. At the end of the process, they give checks to test the integrity of the strings in each database being searched.
The question framework will have identified the initial terms representing each concept for the search string. Because researchers in different contexts use a variety of terms to refer to the same concept, a search string needs to capture all the appropriate varieties of terms (including common misspellings) to represent each concept. If they fail to do this, they risk missing relevant studies from the review. For example, if the original term identified was poultry, the search string could expand to poultry OR fowl OR chicken* OR chukar* OR duck* OR emu OR emus* OR goose OR geese OR "guinea fowl*" OR ostriche* OR quail* OR rea OR rheas OR turkey* OR waterfowl.
Subject focused databases are built around a thesaurus of subject specific terms that are applied to each record to enhance discoverability of the research. These terms, called controlled vocabulary, are added to each record. By adding additional, regularized terms which
capture the main concepts and topics addressed in the research the record represents, these terms improve a searcher's ability to find relevant research.
A comprehensive search in a database with controlled vocabulary combines free text search terms with controlled vocabulary terms for each concept. Identifying the appropriate controlled vocabulary requires spending time exploring the thesaurus (in PubMed and Medline called MeSH headings) for each database. Since terms can have different meanings, shaped by the context of each database's subject focus, attention must be paid to the scope notes and the hierarchy surrounding each term. These both help the searcher understand the assigned to a term within that database.
Scope note for Eggs in MeSH, the National Library of Medicine controlled vocabulary thesaurus:
Scope note for Eggs in the FSTA thesaurus:
Controlled vocabulary terms can also be what’s called exploded. An exploded term will search the term itself, plus the narrower, more specific, terms below it in thesaurus hierarchy.
If you were to explode the terms eggs in the FSTA thesaurus, the search would then include EGGS, DRIED EGGS, DUCK EGGS, PIDAN, EGG SHELL MEMBRANES, EGG SHELLS, EGG WHITES, EGG WHITES LYSOZYMES, EGG YOLKS, EGGS LYSOZYMES, GOOSE EGGS, HAUGH SCORE, and QUAIL EGGS.
In Mesh, exploded eggs would include Eggs, Egg Proteins, Dietary, Egg White and Egg Yolk.
If we turn back to our poultry search, the exploded controlled vocabulary term from could be added to the search term like this:
Search syntax refers to how search terms are connected and sometimes modified to optimise a search. Although there are some common across searches exactly how to use them is driven by the platform. Errors in syntax on a platform will cause a search not to run correctly or not to run at all.
As terms are identified, they need to be built into a search string using the Boolean operators OR and AND. As demonstrated above, the OR Boolean operator is placed between each of the terms that represent the same concept. The AND Boolean separates each concept group from the others. It is crucial to master the syntax of each database to build the search correctly. To learn more about Boolean operators, see the IFIS guide to literature searching.
A search combining two concepts, poultry and salmonella, could look like this:
Good practice point: Do not use the NOT Boolean operator in the final systematic review search string. NOT eliminates records with the term indicated, and risks eliminating relevant records in which the terms were not configured as anticipated. Not is extremely useful for determining and testing each term included in the string, but the final string should be built with blocks of terms representing each concept connected with OR, and then each block connected with AND.
In many databases terms can be truncated. Using the symbol dictated by the database's platform, a symbol can be inserted to represent the possibility of more characters following on from the characters typed. For instance typing toxin* would return both toxin and toxins, but also results with toxinovorans, toxinotype,toxinotypes, toxinotyping, toxinogenisis, etc.
Line 4 of the search above could be simplified to salmonell*
Some databases include the option to limit the number of characters that follow a truncation symbol. When that is possible, typing adult$2, for instance, would return results with adult and adults, but not adulteration.
In cases where limiting the number of cases is not possible but would be useful, extraneous results can be screened out at that stage of the process. Alternatively, rather than truncating a term, typing out both, or all, exact acceptable forms of the term in the search string is an option. Typing out adult OR adults, which would not include records with adulteration in the results.
Sometimes two words together will represent a concept, and will often be stated together as a phrase, with the two words next to each. Rather than searching such words as a phrase encased in quotation marks, proximity operators should be used. A proximity operator dictates that two terms must fall within a certain number of words of each other.
For example, lactic acid might seem like a phrase that would always appear as a phrase in the literature, but in practice there is research that refers to lactic and acetic acid, to lactic and volatile acids, or, at times, to acid lactic. A search string built using a phase instead of a proximity operator would miss those records.
To increase the likelihood that a search term will not inadvertently exclude relevant records, choosing the right form for each term is crucial. Rules of thumb:
Testing the variations that you consider using for each term is a crucial component of building your search.
Sometimes a review will only be interested in a single study type, such as randomized control trials. Or are review may only consider human studies. Although in some databases there is a tickbox option to filter results for human studies, those tend to be prone to some error and should be avoided. Cornell University Libraries have collected links to a wide variety of filters on their Guide to Evidence Synthesis.
Because of the vagaries of language and reporting, it is often a safer approach to include the necessary study type as a screening element rather than applying it to the literature search.
Language restrictions risk introducing bias into a systematic review and skewing its conclusion. The amount of risk varies across disciplines.  Researchers examining conservation ecology evidence syntheses suggest that including non-English language studies can enhance the evidence base "with specific types of scientific information that is not available in English-
language studies, especially in disciplines dealing with more geographically and taxonomically diverse targets and phenomena than healthcare". 
No known research has been conducted specifically on the risks of language bias systematic reviews in the food sciences. However, a significant proportion of the literature is written in languages other than English, and excluding those studies, if they otherwise meet the screening criteria for the review, may introduce a risk of bias.
Many databases’ records provide an English-language summary sufficient to conduct the first level of screening. If the review team lacks the language capacity to score any studies that pass through to the next level of screening, they can consider outreach to develop collaborations.  When that is not possible, good practice is to include the studies in the review's reporting as awaiting classification rather than as excluded studies.
Search strings must be tested to ensure that they do not include errors and are comprehensive. The PRESS guideline and checklist outline the elements to check to be sure that a search will perform its function to the standard needed in a systematic review.
1.Translation of the research question:
Assess whether the research question has been correctly translated into search concepts.
2. Boolean and proximity operators:
Assess whether the elements addressing the search question have been correctly combined with Boolean and/or proximity operators.
3. Subject headings (database specific):
Assess whether there is enough scope in the selection of subject headings to optimize recall.
4. Text word search (free text):
Assess whether search terms without adequate subject heading coverage are well represented by free-text terms, and whether additional synonyms or antonyms (opposites) and related terms are needed.
5. Spelling, syntax, and line numbers:
Assess correct use of spelling, correct use of syntax and correct search implementation.
6. Limits and filters:
Assess whether the limits used (including filters) are appropriate and have been applied correctly.
For a clear and concise overview of searching steps for a systematic review, see Livoreil, B, et al. Systematic searching for environmental evidence using multiple tools and sources. Environmental Evidence, 2017. 6(23). https://doi.org/10.1186/s13750-017-0099-6
For guidance on the basic mechanics of constructing a search string see https://ifis.libguides.com/literature_search_best_practice/search-strategy