Introduction

A database search for a systematic or scoping review is part of the review's method, in the same way that a sampling plan is part of a survey. Another team should be able to read the search, judge whether it was likely to find the relevant studies, and run it again. Lesson 3 explained how bibliographic databases index records and return every record that matches a Boolean query. This lesson follows the evidence team of the fictional Cedar Valley Health Authority as its librarian and student intern build, translate, peer review and document the database search for their rapid scoping review.

1.1 From Review Question to Search Concepts

A search concept is one idea from the review question that a relevant record must address, expressed as a group of interchangeable terms. The first decision in building a search is which parts of the question become concepts. Each concept is later joined to the others with the Boolean operator AND, so a record is retrieved only if it contains at least one term from every concept. Every concept that is added therefore removes records, including some relevant ones whose titles and abstracts happen to describe that idea in words the search does not include. For this reason most review searches combine two to four concepts, and the remaining parts of the question are applied by reviewers during screening.

A useful test for each element of the question is to ask whether the authors of a relevant study would almost always describe it in the title, abstract or subject headings. The population and the main problem or intervention usually pass this test. Elements that authors describe inconsistently, or leave out of the abstract, usually fail it. Study design is a common example in scoping reviews, because a scoping review includes many designs and their labels vary. Setting is another, because a paper about a seniors' centre program may never use the word community.

The Cedar Valley team applied this test to the PCC question. The population (adults aged 65 and older) and the outcome or problem (loneliness or social isolation) are described in nearly every relevant abstract, so both became concepts. The concept element (community-based interventions) is more loosely described, but without it the search would retrieve thousands of studies that measure loneliness without testing any intervention, so it became a third concept with a deliberately wide set of terms. The context element (community settings in high-income countries) is judged at screening, and the date criterion is applied as a limit at the end of the search. The team decided against adding words such as evaluation or effect as a fourth concept, because many evaluations describe themselves only by their design or measures, and those records would be lost.

1.2 The Concept Table

A concept table is a grid that lists, for each search concept, the subject headings and the free-text terms that will be searched, with notes on decisions. It makes the logic of the search visible before any syntax is written, gives a peer reviewer something to check, and becomes the starting point for translation into other databases, where the headings change and the free-text terms usually stay the same (Section 3).

Two kinds of terms appear in each row. Subject headings are terms from a database's controlled vocabulary, the fixed list of headings that indexers assign to records to describe what they are about; in MEDLINE this vocabulary is MeSH (Medical Subject Headings). Free-text terms, sometimes called keywords or text words, are words searched in the title, abstract and author keyword fields, so they find records whatever headings were assigned. The table below is the Cedar Valley concept table for MEDLINE.

Concept C deliberately mixes headings for the mechanism of an intervention (social support) with headings for who delivers it (volunteers, peers), because the team cannot predict how each study will be indexed and relies on the AND with concepts A and B to remove irrelevant records. The free-text column lists word forms, which the syntax in Section 2 compresses, for example by searching lonel$ to find lonely, loneliness and lonelier.

1.3 Finding Synonyms and Variants

Most of the work of a concept table lies in the free-text column. A synonym that is missing from the table is a group of relevant records the search cannot find, and nothing in the results will reveal the gap. The JBI approach to scoping review searches begins for this reason with a limited search of at least two databases, followed by analysis of the words in the titles and abstracts of relevant records and of the index terms used to describe them, before the full search is built and run across all databases (Peters et al., 2020). The cards below describe the main sources of terms.

Older terms raise a point about language. A review team may decide, for good reasons, to write about older adults rather than the elderly. The search must still include elderly, because a large share of the literature on this population used that word, and the search has to find the literature as it was written.

Ambiguous terms raise a different point. Isolation and connector both have other meanings that would flood the results with irrelevant records. The usual solution is to keep the term and tie it to a context word with a proximity operator, which Section 2 explains.

1.4 Controlled Vocabulary in Practice

Lesson 3 explained why controlled vocabularies exist: they gather records about the same topic under one heading, whatever words the authors used. In practice, a subject heading search behaves in ways that a searcher must understand before using it. The items below describe the features that matter most when building a review search in MEDLINE, and the same ideas apply to Emtree in Embase, CINAHL Subject Headings and the APA Thesaurus of Psychological Index Terms.

These features explain the standard rule for review searches: each concept is searched with both subject headings and free text, combined with OR. Headings find records that describe the concept in unexpected words, and free text finds records that are unindexed or indexed under unexpected headings.

1.5 Testing the Draft with Known Studies

A test set (also called a validation set or benchmark set) is a group of studies the team already knows to be relevant, gathered from a preliminary search, from existing reviews or from experts. Once a draft strategy is written, the librarian checks how many of the test-set studies it retrieves in each database where they are indexed. A study that is indexed in the database but missed by the strategy reveals a gap: a synonym that is missing, a heading that was not used, or a concept that is too narrow.

The Cedar Valley team assembled a test set of ten evaluations of community programs for lonely or isolated older adults. Nine of them were indexed in MEDLINE, and the first draft of the MEDLINE strategy retrieved eight. The missing study described its outcome only as social connectedness and was indexed under headings that the draft did not include. The intern added a line for social connection and connectedness to concept A, and the revised strategy retrieved all nine. The team also checked whether concept C was costing recall by running concepts A and B alone. That search also retrieved all nine studies, together with many more records that measured loneliness without testing an intervention, so the team kept concept C.

A test set shows whether the search finds studies like those the team already knows, and it cannot show that the search finds studies described in ways nobody anticipated. Studies used to choose the terms will be retrieved almost by construction, so a test set is more informative when some of its studies were set aside before the terms were chosen. It is one check among several, alongside the peer review in Section 4.

Introduction

The concept table in Section 1 lists what to search. This section turns it into a working search in Ovid, the platform the Cedar Valley librarian uses for MEDLINE. Ovid keeps a numbered search history in which each line can be a set of terms or a combination of earlier lines, and its syntax uses a small set of symbols and operators. The section explains each element and then presents the full strategy. Section 3 shows how the same strategy is written for other databases and platforms.

2.1 Boolean Operators

Boolean operators tell the database how to combine sets of records. OR retrieves records that contain any of the terms it joins, so it widens a search and is used inside a concept. AND retrieves records that contain at least one term from each set it joins, so it narrows a search and is used between concepts. NOT removes every record in the second set from the first, so it narrows a search by exclusion.

NOT needs care, because it removes a record if the excluded term appears anywhere in the searched fields, even when the record is relevant. Suppose the intern wanted to remove studies of children and added not child$.ti,ab. to the search. The line would remove every intergenerational program in which older adults spend time with children, which are among the interventions the review is looking for. The animal-studies line in the table is a safer use of NOT. It first finds records indexed with animal headings that are not also indexed with the heading Humans, and only those records are removed, so a study that involves both people and animals stays in the results. This form of the line is widely used and appears in Cochrane guidance on searching (Lefebvre et al., 2023).

Parentheses and numbered lines

When a single line mixes operators, the database applies its own order of precedence, and that order differs between platforms. The line lonely or isolated and seniors could be read as lonely or (isolated and seniors), which retrieves every record about loneliness at any age, or as (lonely or isolated) and seniors, which is what the searcher intended. Parentheses remove the ambiguity: (lonely or isolated) and seniors. Any line that mixes operators should use them.

Review searches go further and build the strategy in numbered lines, usually one line per subject heading or per group of related free-text terms, followed by a line that combines each concept with OR and a final line that combines the concepts with AND. In Ovid, or/1-6 combines lines 1 to 6 with OR. Numbered lines make errors easier to find, show how many records each line contributes, and let a peer reviewer check every combination. This layout is often called the building-block approach.

2.2 Truncation and Wildcards

Truncation searches every word that begins with a given stem. In Ovid, the dollar sign ($) and the asterisk (*) both act as unlimited truncation, and this lesson uses the dollar sign in Ovid strategies. volunteer$ finds volunteer, volunteers, volunteered and volunteering. A number after the symbol limits the number of extra characters, so club$1 finds club and clubs and stops there.

A wildcard stands for a character inside a word. In Ovid, # stands for exactly one character (wom#n finds woman and women), and ? stands for one character or none (behavio?r finds behavior and behaviour, and neighbo?rhood finds both spellings of neighbourhood). Section 3 shows that other platforms assign different meanings to these symbols, which is a common source of translation errors.

The main risk with truncation is a stem that is too short. The table lists examples the Cedar Valley team considered and rejected.

2.3 Phrase Searching

A phrase search retrieves words only when they appear next to each other in the order given. Ovid treats words typed side by side as a phrase, so later life.ti,ab. searches the phrase later life. Quotation marks are optional in Ovid and are needed mainly when a phrase contains a word Ovid would otherwise read as an operator, such as and or or. Truncation can be used inside a phrase, so friendly visit$ finds friendly visit, friendly visits and friendly visiting. Phrases are precise, but they miss the same idea expressed with the words in another order or separated by other words. Social isolation as a phrase misses isolated socially and social and emotional isolation. Proximity operators address that problem.

2.4 Proximity Operators

A proximity operator retrieves two terms when they appear within a set distance of each other. In Ovid, adj on its own finds terms next to each other in the order typed, and adj followed by a number finds terms within that many words of each other in either order. The number counts the words themselves, so adj3 allows up to two words in between and adj2 allows one. Figure 2 shows how line 4 of the Cedar Valley strategy treats four phrases.

The distance is a judgement about recall and precision. A small number behaves almost like a phrase. A large number finds more variants but also pairs words that belong to different ideas, such as social at the end of one clause and isolation at the start of the next. Distances of two to five words are common in review searches, and the choice should be tested by looking at a sample of the records each setting adds. Proximity also offers a more precise alternative to AND inside a line: (intergeneration$ adj3 program$) requires the two words to be close together, whereas intergeneration$ and program$ retrieves any record that uses both words anywhere.

2.5 Field Tags

A field tag restricts a term to one or more parts of the record. In Ovid, field tags follow the term, between full stops. Searching free-text terms in the title, abstract and author keywords is the usual choice for reviews, because these fields describe what the study is about, whereas the full record also contains author addresses, journal names and reference lists in some databases.

Because .tw. and .mp. cover different fields in different databases, the Cedar Valley librarian lists the fields explicitly with .ti,ab,kf.. Explicit tags make the strategy easier to report, easier to peer review and easier to translate, since the translator can see exactly which fields each line searched.

2.6 The Full MEDLINE Strategy

The strategy below is the final Cedar Valley MEDLINE search, after the test-set check in Section 1 and the peer review described in Section 4. It is shown as it would be exported from Ovid, which is the form required for reporting. The search date and record count belong to the fictional project.

Database: Ovid MEDLINE(R) ALL <1946 to February 09, 2026>
Date searched: 10 February 2026
------------------------------
1   Loneliness/
2   Social Isolation/
3   lonel$.ti,ab,kf.
4   ((social$ or emotional$) adj3 (isolat$ or disconnect$ or exclu$ or withdraw$)).ti,ab,kf.
5   ((isolated or isolation) adj3 (older or elder$ or senior$ or people or person$)).ti,ab,kf.
6   (social$ adj2 connect$).ti,ab,kf.
7   or/1-6
8   exp Aged/
9   (older adj3 (adult$ or people or person$ or individual$ or men or women or resident$ or population$ or patient$)).ti,ab,kf.
10  (elder$ or senior$ or geriatric$ or gerontolog$ or octogenarian$ or nonagenarian$ or centenarian$).ti,ab,kf.
11  (old age or later life or late life or retire$).ti,ab,kf.
12  or/8-11
13  Social Support/
14  Social Participation/
15  Community Networks/
16  Volunteers/
17  Peer Group/
18  Intergenerational Relations/
19  (social$ adj2 prescri$).ti,ab,kf.
20  ((community or link or neighbo?rhood or care) adj2 (connector$ or navigator$ or worker$)).ti,ab,kf.
21  (befriend$ or friendly visit$).ti,ab,kf.
22  ((peer or volunteer$ or lay) adj3 (support$ or visit$ or mentor$ or companion$)).ti,ab,kf.
23  (intergeneration$ adj3 (program$ or activit$ or project$ or contact$)).ti,ab,kf.
24  ((group or groups or club or clubs or class or classes) adj3 (social$ or leisure or recreation$ or art or arts or exercis$ or walking or garden$ or music or singing or choir$)).ti,ab,kf.
25  (communit$ adj3 (intervention$ or program$ or service$ or initiative$)).ti,ab,kf.
26  or/13-25
27  7 and 12 and 26
28  exp animals/ not humans/
29  27 not 28
30  limit 29 to yr="2010 -Current"

Line 30 retrieved 742 records in the fictional search. A real export also lists the number of records for every line, and those counts are useful in peer review because a line that retrieves nothing, or an implausibly large number, often signals an error.

Introduction

No single database indexes all the literature on a health question, so reviews search several, and the strategy developed in one database, the master strategy, is rewritten for each of the others. This search translation changes both the controlled vocabulary and the syntax. This section translates the Cedar Valley MEDLINE strategy and then considers search filters and limits.

3.1 Why Search Several Databases

Databases differ in the journals and record types they cover, as the cards below summarize. For Cochrane reviews of interventions, the Cochrane Handbook treats searches of the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE and Embase as the minimum, with other databases added according to the topic (Lefebvre et al., 2023).

Research on loneliness among older adults appears in gerontology, nursing, psychology, social work and public health journals, so the Cedar Valley team planned searches of all five databases below and a separate grey-literature search for program reports (Lesson 5).

3.2 Databases and Platforms

A database is a collection of records with its own coverage and, usually, its own thesaurus. A platform (sometimes called an interface or host) is the software used to search it. The same database can be searched on different platforms: MEDLINE through PubMed or Ovid, Embase through Embase.com or Ovid, and APA PsycInfo through Ovid or EBSCOhost. The controlled vocabulary belongs to the database and the syntax belongs to the platform, so translating from Ovid MEDLINE to Ovid Embase changes the headings and keeps most of the syntax, whereas translating to CINAHL on EBSCOhost changes both. The platform must also be reported, because the same strategy can retrieve different records on different platforms.

3.3 What Changes in Translation

The table compares the main syntax elements on four platforms. Each platform's help pages remain the authority, since syntax is revised over time.

Three kinds of difference cause most translation errors. Ovid and EBSCOhost give opposite meanings to # and ?, so a copied wildcard searches something different. In Web of Science, $ stands for one character or none, so the Ovid term lonel$ copied unchanged would miss loneliness entirely. Platforms also count proximity differently: Ovid counts the search words themselves, so adj3 allows two words between them, whereas EBSCOhost and Web of Science count only the words in between, so the equivalents are N2 and NEAR/2.

3.4 Translating the Cedar Valley Strategy

The translation proceeds line by line. For each subject heading, the librarian reads the target thesaurus's scope note and tree and chooses the closest heading, which may differ in name or scope. For each free-text line, the librarian changes the field tags, truncation, wildcards and proximity operators. The table records the heading decisions, and the tabs show the translated strategies.

All three vocabularies are revised annually, so headings must be checked in each thesaurus at the time of searching. When no heading matches closely, the librarian relies on the free-text lines and records the decision.

Database: CINAHL (EBSCOhost)
Date searched: 10 February 2026
S1   (MH "Loneliness")
S2   (MH "Social Isolation")
S3   TI lonel* OR AB lonel*
S4   TI ( (social* OR emotional*) N2 (isolat* OR disconnect* OR exclu* OR withdraw*) ) OR AB ( (social* OR emotional*) N2 (isolat* OR disconnect* OR exclu* OR withdraw*) )
S5   TI ( (isolated OR isolation) N2 (older OR elder* OR senior* OR people OR person*) ) OR AB ( (isolated OR isolation) N2 (older OR elder* OR senior* OR people OR person*) )
S6   TI ( social* N1 connect* ) OR AB ( social* N1 connect* )
S7   S1 OR S2 OR S3 OR S4 OR S5 OR S6
S8   (MH "Aged+")
S9   TI ( older N2 (adult* OR people OR person* OR individual* OR men OR women OR resident* OR population* OR patient*) ) OR AB ( older N2 (adult* OR people OR person* OR individual* OR men OR women OR resident* OR population* OR patient*) )
S10  TI ( elder* OR senior* OR geriatric* OR gerontolog* OR octogenarian* OR nonagenarian* OR centenarian* ) OR AB ( elder* OR senior* OR geriatric* OR gerontolog* OR octogenarian* OR nonagenarian* OR centenarian* )
S11  TI ( "old age" OR "later life" OR "late life" OR retire* ) OR AB ( "old age" OR "later life" OR "late life" OR retire* )
S12  S8 OR S9 OR S10 OR S11
S13  (MH "Support, Psychosocial+")
S14  (MH "Volunteer Workers")
S15  (MH "Peer Group")
S16  (MH "Intergenerational Relations")
S17  TI ( social* N1 prescri* ) OR AB ( social* N1 prescri* )
S18  TI ( (community OR link OR neighbo#rhood OR care) N1 (connector* OR navigator* OR worker*) ) OR AB ( (community OR link OR neighbo#rhood OR care) N1 (connector* OR navigator* OR worker*) )
S19  TI ( befriend* OR (friendly N1 visit*) ) OR AB ( befriend* OR (friendly N1 visit*) )
S20  TI ( (peer OR volunteer* OR lay) N2 (support* OR visit* OR mentor* OR companion*) ) OR AB ( (peer OR volunteer* OR lay) N2 (support* OR visit* OR mentor* OR companion*) )
S21  TI ( intergeneration* N2 (program* OR activit* OR project* OR contact*) ) OR AB ( intergeneration* N2 (program* OR activit* OR project* OR contact*) )
S22  TI ( (group OR groups OR club OR clubs OR class OR classes) N2 (social* OR leisure OR recreation* OR art OR arts OR exercis* OR walking OR garden* OR music OR singing OR choir*) ) OR AB ( (group OR groups OR club OR clubs OR class OR classes) N2 (social* OR leisure OR recreation* OR art OR arts OR exercis* OR walking OR garden* OR music OR singing OR choir*) )
S23  TI ( communit* N2 (intervention* OR program* OR service* OR initiative*) ) OR AB ( communit* N2 (intervention* OR program* OR service* OR initiative*) )
S24  S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23
S25  S7 AND S12 AND S24
S26  S25   Limiters - Published Date: 20100101-20261231

Database: Embase <1974 to 2026 February 09>
Date searched: 10 February 2026
1   loneliness/
2   social isolation/
3   lonel$.ti,ab,kw.
4   ((social$ or emotional$) adj3 (isolat$ or disconnect$ or exclu$ or withdraw$)).ti,ab,kw.
5   ((isolated or isolation) adj3 (older or elder$ or senior$ or people or person$)).ti,ab,kw.
6   (social$ adj2 connect$).ti,ab,kw.
7   or/1-6
8   exp aged/
9   (older adj3 (adult$ or people or person$ or individual$ or men or women or resident$ or population$ or patient$)).ti,ab,kw.
10  (elder$ or senior$ or geriatric$ or gerontolog$ or octogenarian$ or nonagenarian$ or centenarian$).ti,ab,kw.
11  (old age or later life or late life or retire$).ti,ab,kw.
12  or/8-11
13  social support/
14  volunteer/
15  peer group/
16  (social$ adj2 prescri$).ti,ab,kw.
17  ((community or link or neighbo?rhood or care) adj2 (connector$ or navigator$ or worker$)).ti,ab,kw.
18  (befriend$ or friendly visit$).ti,ab,kw.
19  ((peer or volunteer$ or lay) adj3 (support$ or visit$ or mentor$ or companion$)).ti,ab,kw.
20  (intergeneration$ adj3 (program$ or activit$ or project$ or contact$)).ti,ab,kw.
21  ((group or groups or club or clubs or class or classes) adj3 (social$ or leisure or recreation$ or art or arts or exercis$ or walking or garden$ or music or singing or choir$)).ti,ab,kw.
22  (communit$ adj3 (intervention$ or program$ or service$ or initiative$)).ti,ab,kw.
23  or/13-22
24  7 and 12 and 23
25  (exp animal/ or nonhuman/) not exp human/
26  24 not 25
27  limit 26 to yr="2010 -Current"

#1  'loneliness'/de OR 'social isolation'/de
#2  lonel*:ti,ab,kw
#3  ((social* OR emotional*) NEAR/3 (isolat* OR disconnect* OR exclu* OR withdraw*)):ti,ab,kw
#4  ((isolated OR isolation) NEAR/3 (older OR elder* OR senior* OR people OR person*)):ti,ab,kw
#5  (social* NEAR/2 connect*):ti,ab,kw
#6  #1 OR #2 OR #3 OR #4 OR #5

1  Loneliness/
2  Social Isolation/
3  lonel$.ti,ab,id.
4  ((social$ or emotional$) adj3 (isolat$ or disconnect$ or exclu$ or withdraw$)).ti,ab,id.

#1  TS=(lonel* OR ((social* OR emotional*) NEAR/2 (isolat* OR disconnect* OR exclu* OR withdraw*)) OR ((isolated OR isolation) NEAR/2 (older OR elder* OR senior* OR people OR person*)) OR (social* NEAR/1 connect*))

Translation tools and checks

The Polyglot Search Translator converts the syntax of a strategy between common platforms, and a randomized trial found that it reduced the time searchers spent on translation (Clark et al., 2020). Such tools leave the choice of headings to the searcher, so every translated heading still has to be checked. After translation, the librarian runs each line, compares its record count with the corresponding MEDLINE line, investigates any line that retrieves nothing or far more than expected, and repeats the test-set check in each database.

3.5 Validated Search Filters

A search filter (also called a hedge) is a ready-made block of search lines that retrieves a type of record, most often a study design such as randomized controlled trials, and sometimes an age group or geographic area. A validated filter has been tested against a reference set of records of known status, so its sensitivity (the share of relevant records it retrieves) and precision are reported, often in versions that favour one or the other. A filter is designed for one database and platform, and it should be used as published, cited and, if modified, described as modified.

3.6 Limits

A limit restricts results by a property of the record, such as language, publication year, publication type or indexed age group. Every limit removes records, so each needs a reason that follows from the protocol, and the reason must be reported. Language limits can bias a review toward studies published in English, and Cedar Valley applies none, because its protocol makes English and French reports eligible and asks the team to list relevant reports in other languages. The date limit follows the protocol's criterion of publication from January 2010 onward.

Limits that rely on indexing have a further weakness. Ovid's age-group limits depend on the age headings assigned to each record, so they miss every unindexed record, which is why concept B searches the population with headings and free text instead. The animal-studies lines act as a simple filter and are reported with their source.

Introduction

A search strategy can contain errors that no one notices, because a missing synonym or a wrong line number produces results that look reasonable. The two practices in this section address that problem from different directions. Peer review checks the strategy before it is run, while there is still time to fix it. Reporting describes the strategy after it is run, so that readers can judge it and other teams can reproduce it. Figure 4 places both practices in the sequence of work covered in this lesson.

4.1 Why Searches Are Peer Reviewed

Studies of the search strategies published with systematic reviews have found errors of several kinds, including missing synonyms, missing or unexploded subject headings, incorrect line combinations, spelling mistakes and misused operators, and many of these errors could reduce recall (Sampson & McGowan, 2006). An error of this kind leaves no trace in the results, since a search that misses a set of relevant records still returns plenty of records. The protocol and the manuscript of a review are peer reviewed as a matter of routine, and search peer review applies the same logic to the step on which every later step depends. It is most useful before the searches are run, because a strategy corrected afterwards means rerunning every database and screening again.

4.2 The PRESS 2015 Guideline

PRESS (Peer Review of Electronic Search Strategies) is a guideline for the peer review of search strategies by a second searcher, usually another librarian or information specialist. Its current version, the PRESS 2015 Guideline Statement, was developed from a systematic review of the evidence on search errors, a survey of searchers and a consensus process, and it includes an evidence-based checklist organized in six elements (McGowan et al., 2016). For each element, the reviewer records comments and rates the strategy as needing no revisions, suggested revisions or required revisions, and then gives an overall rating. Reviewers usually work on the master strategy, because errors found there would otherwise be copied into every translation.

4.3 A Worked PRESS Review

Before the final MEDLINE strategy shown in Section 2 existed, the intern's first complete draft was sent to a second librarian at the university for peer review. The draft is shown below with the reviewer's comments, organized by PRESS element.

Draft 1, Ovid MEDLINE(R) ALL, submitted for peer review
1   Loneliness/
2   (lonliness or lonely).tw.
3   (social$ adj3 isolat$).tw.
4   or/1-3
5   Aged/
6   (older adult$ or elderly or senior$).mp.
7   5 or 6
8   (social prescri$ or befriend$ or volunteer$).tw.
9   ((group or club) adj3 (art$ or exercise or music)).tw.
10  (communit$ adj3 program$).tw.
11  or/8-9
12  (evaluat$ or effect$ or outcome$).tw.
13  4 and 7 and 11 and 12
14  13 not (child$ or adolescen$).tw.
15  limit 14 to (english language and "aged (65 and over)")

The overall rating was revisions required. The intern revised the strategy, wrote a short response that listed each comment and the change made (or the reason for making no change), and returned both to the reviewer, who confirmed the revisions. The test-set check from Section 1 was then repeated on the revised strategy. The result is the strategy in Section 2.6, which was then translated for the other databases.

Writing useful feedback

The worked review illustrates the qualities that make PRESS feedback useful. Each comment names a line number, explains what the problem would do to the results, and proposes a correction. Required revisions are separated from suggestions, so the searcher knows which changes affect the validity of the search and which are matters of judgement. The reviewer also respects decisions that the searcher made for documented reasons, such as leaving context out of the search, and asks about them before recommending a change. Student peer review in this course uses the same structure.

4.4 Reporting with PRISMA-S

PRISMA-S is a sixteen-item extension of the PRISMA Statement that sets out what a review should report about its literature searches (Rethlefsen et al., 2021). It is used alongside PRISMA 2020 (Page et al., 2021) and its other extensions, including PRISMA-ScR for scoping reviews, which Lesson 10 discusses. The items fall into four groups. Several of them concern searches outside bibliographic databases, which Lesson 5 covers, and the last two concern record management, which Lesson 7 covers.

Keeping a search log

A team can report these items only if it records them while searching, because platforms change and several details cannot be recovered later. A search log is a running record, kept for every search, of the database and platform, the coverage label the platform displays (for example, 1946 to February 09, 2026), the date, the person who ran it, the full strategy with the number of records for each line, the limits applied, the name of the exported file, and notes on anything unusual. Ovid and EBSCOhost can export the search history with line counts, and saving that export on the day of the search is the simplest way to preserve the strategy exactly as run. Some teams also deposit their strategies in an open archive such as searchRxiv so that others can reuse them.

A reporting example

The table and paragraph below show how the Cedar Valley database searches will be reported in the methods section of the scoping review. The record counts are those of the fictional project, and the five databases together retrieved 2,480 records.

The paragraph covers items 1, 2, 8, 9, 10, 11, 13, 14 and 15, and the appendix supplies the strategies required by item 8. Item 16 is reported in the record-management section of the review, where Lesson 7 shows how the 610 duplicates are removed to leave 1,870 records for screening, and items 3 to 7 are reported with the grey-literature methods.

Updating searches

Searches age. Cochrane's conduct standards for intervention reviews expect searches to be rerun within twelve months before a review is published, and many other review teams follow similar practice. The Cedar Valley brief is due within twelve weeks, and no update is planned in that time. The team saves every strategy on the platform where it was run, so the searches can be rerun quickly if the review is later published or updated. Any update is reported under PRISMA-S item 12 with its date and method.

Bringing It All Together

This lesson followed the Cedar Valley evidence team as it turned a review question into a database search that another team could judge and reproduce. The work began with a concept table, which decided which parts of the question to search and which to apply at screening, and listed subject headings and free-text synonyms for each concept. A test set of known studies showed where the draft fell short. The concepts were then written in Ovid syntax as numbered lines that use Boolean operators, truncation, wildcards, adjacency and explicit field tags, with limits that follow from the protocol.

The master strategy was peer reviewed with the PRESS 2015 guideline before it was translated, so that corrections carried into every database. Translation changed both the vocabulary, because MeSH, Emtree, CINAHL Subject Headings and the APA Thesaurus differ, and the syntax, because Ovid, EBSCOhost, Embase.com and Web of Science use different symbols and count proximity differently. A search log kept during searching allowed the team to report its five database searches according to PRISMA-S, with full strategies, dates, limits and record counts.

The final reflection asks you to apply the PRESS elements and PRISMA-S to a colleague's draft search.

Final Knowledge Assessment

Terms, tools and people introduced in this lesson on building and documenting database searches.