Because searches conducted for systematic reviews must be as exhaustive as possible, we not only need to do controlled vocabulary searches, but also regular free-text searches, or sometimes called "Natural Language" searches. This will catch anything that's not properly indexed in databases due to a variety of reasons. Indicating a free-text search is pretty simple in PubMed - you just have to put quotes around your search terms. This will get around the automatic term mapping layer and search the entered term as-is in all fields. We can see the actual query here in the "Search Details" box. This quotes trick also works for "Phrase Searching", that is, if you put in multiple words and want to search them exactly as they spelled, spaced, and ordered. So if i search for "myocardial infarction"... ...but if I search for "miocardial infarction", with an "i" I can still get some results - these would be the articles where the term is, for some reason, spelled in that way. And if I search for "infarction myocardial" I can still get a result set where the phrase "infarction myocardial" actually appears. You will notice that there is a TW "Text Words" field here, and it includes all words and numbers in these fields - title, abstract, et cetera. You'll notice that it includes most, but not all fields - especially the author field. An obvious disadvantage of free text, or natural language searching, is that it does not handle the "synonym" problem at all, so if you search for "adolescent" it does not automatically find "young person", "child", "teenager", "youth", et cetera. And that's where our controlled vocabulary search comes in handy. But if the database you are using does not support a controlled vocabulary system (and there are many of them, such as the Web of Science or Scopus), you would have to think of all the possible synonyms and put them into the search box. Our concept table, introduced in the previous video, includes all the synonyms you can think of for your concepts. Even though a database does support controlled vocabulary, in the case of a systematic review search, you are required to be absolutely exhaustive in your search, so you should also do a free-text search to compliment your controlled vocabulary search - to catch anything that's not properly indexed by the database. The other disadvantage of free-text searching is that it does not automatically handle spelling variations. For example, if you search for "organization" with a "z", it will not find "organisation" with an "s". If you search for "child" as a free-text word, it does not automatically find "children" in the plural. And if you search for "cardiology", it does not automatically match all of its inflectional forms, such as "cadiologic" and "cardiologist", and so forth. Fortunately, many databases offer features such as "truncation" and "wild card" searching to help solve this problem, even though not all databases support all truncation and wild card features. In PubMed, only truncation is supported. To "truncate" a search term, that is, to search for all terms that begin with a string, enter the string followed by an asterisk. So the example here is that if you put in flavor asterisk, you will not only find "flavor", but also "flavored", "flavorful", "flavoring", et cetera. Boolean logic is an important concept in database searching, which helps you combine your terms appropriately. Nearly all databases have some sort of Boolean operator support, even though they may support them differently. To put it simply, the Boolean "AND" creates an intersection of the search terms connected by it, and the Boolean "OR" creates a union of the terms. Let's illustrate this using PubMed as an example. So if I want all articles on the subject "myocardial infarction" that was published in year 2012, I can say "myocardial infarction MeSH terms AND 2012 date of publication". Now, it is recommended that you spell the Boolean operators with capital letters in PubMed, so that it's easier to identify them in the search strategy. It is no longer required though, because the automatic term mapping layer, in most cases, can automatically map lower case Boolean operators to uppercase ones. The Boolean "OR" is best used to connect a series of synonymous terms to create a union of results representing a single concept. For example, "hypertension MeSH terms OR high blood pressure text words OR hypertens asterisk text words". Another example, "angio tension inhibitor MeSH terms OR ace inhibitor text words". Here's an example to combine the two. Notice that the parentheses indicate the precedence of execution in Boolean operations. While not commonly used, the Boolean "NOT" subtracts results that match the term that follows it from the results that match the term that precedes it. So, if I want all the previous search results, but not anything that is a letter in publication type, I can say "NOT letter pt". Normally it is recommended that you use "NOT" very, very cautiously, because you could inadvertently omit a lot of search results that could be potentially relevant. Boolean operators are more commonly used to combine multiple search queries in your search history, so if I switch to the advanced view, I can see my search history in PubMed and you will see that every search that I ran was given a set number that starts with a pound sign. And let's say I want to combine these two with "AND", that is, to find the overlap of these searches... I can just say... ...and you can see that this exactly the same as the search where I explicitly spelled out my search terms. So you could use Boolean operators to combine search terms, as well as search queries. A rule of thumb is that you should put all terms about the same concept in one query, and then you combine the queries that are about different concepts. So in our example here, all the concepts about hypertension are here, and everything about ace inhibitors here, and then I can combine them using Boolean operators. Another common technique for further narrowing your search, is to use limits and filters. Common limits or filters include age groups, publication date ranges, species - human or animal, languages, publication type, reviews, clinical trials, cohort studies, et cetera. In PubMed these limits are called "filters". They are typically listed dynamically to the left of the search results. Now, this is only a partial list of all the filters available in PubMed. To see all filters, click here. So I'm going to put "Age" and "Language" in here... ...and those filters showed up. By default, under each filter category, only a limited number of filter items will show up. If a "More" link is available, you can click on it, and see more items. For publication dates, instead of a "More" link, you get a "Custom Range" so you can fill out this and choose a date range. Now these only change the display of filters - to actually apply the filter to the search result, you have to actually click on it. So this will give you all the English language results, and this will add a human filter on top of the English filter - you can see that there is a check mark beside each of the applied filters. You can clear the filters in each category using the "clear" links, and you can clear all filters using the "Clear All" links. So in this video, we went over some basic concepts in building search strategies, such as field qualifications, controlled vocabulary search versus free text natural language search, truncation and wild card searching, Boolean logic, limits and filters. We've been using PubMed as an example, which does not support all the search features available, such as wild card searching, and adjacency searching. In the next video, I'm going to use OvidSP, Scopus, and the Web of Science as examples to further explore database searching techniques. I'll see you next time.