 PubMed does a lot of things behind the scenes to retrieve relevant search results, even when you conduct the most basic of searches. Quite often this means that you can quickly type a few terms into the search box and find articles on your topic. But when these initial results are not retrieving what you need, it's important to know what PubMed is doing so you can adjust the search to make it more relevant. In this lecture we'll discuss the automatic term mapping and best match features, two default features that PubMed applies to its search automatically. We'll look at what these features are, why they're valuable, and what to do when they are not providing you with the results that you need. For this discussion we'll use the following research question. Given the current opioid epidemic, you want to learn more about prescribing patterns related to opioids with a specific interest in using electronic medical records as a way to monitor trends in the opioids prescribed. You've probably noticed that when you conduct the basic search in PubMed, you get a lot of results. Typing opioids alone yields over 170,000 results. Opioids and prescribing patterns brings us down to 3500 results, and adding electronic medical records gives us 121 results. A smaller number, but still a lot to manage if we're only looking for a handful of articles. The fact that PubMed is one of the largest biomedical databases contributes to that large number of results, but automatic term mapping is also a factor. To better understand what automatic term mapping is, let's go to the PubMed search details. First, click on the advanced link underneath the search box. Then, scroll down to the history and search details box. Now, click on the arrow underneath the details heading. Here, we can see everything that PubMed did behind the scenes when we ran our search. Looking at the search details can be a bit confusing, but the important thing to note is that PubMed does a lot more than simply look for articles that use the words electronic medical records and prescribing patterns. With automatic term mapping, PubMed automatically takes every concept entered in the search box and searches it against PubMed's many indexes. First, PubMed looks for a match in its subject translation table, which includes its index of medical subject headings commonly referred to as mesh. If PubMed finds a corresponding mesh, then it automatically adds this mesh to the search and stops mapping. In our search scenario, PubMed matched our first two terms to their corresponding mesh, electronic health records, and analgesics, opioids. It also mapped analgesics, opioids to its corresponding pharmacological action term. Every drug and chemical mesh heading has been assigned one or more heading that describe its pharmacological action. If PubMed doesn't find a mesh, then it moves on to the journals index, authors index, and then the investigators index in that order. If it finds a match in one of these indexes, the match is added to the search and the mapping stops. If no match is found, which is the case with prescribing patterns, then PubMed breaks apart the phrases in the search and tries to map the individual words to these indexes. While PubMed wasn't able to map the word prescribing to a mesh, it mapped patterns to the mesh term behaviors. The last step that PubMed does if a single word can't be mapped is that it takes the single words, combines them with and, and searches in all fields except for the place of publication and any date fields such as the date that the item was created or indexed. Regardless of whether PubMed finds a match or not, it also automatically adds additional British and American spellings, singular and plural word forms, and other synonyms to the search. This step explains a lot of the additional search terms in the search details. For example, you can see that PubMed automatically added singular and plural variations of the word prescribing. In many cases, automatic term mapping is helpful for finding relevant articles without developing a sophisticated search. From our basic search, we were able to get 121 results, which can act as a good starting place for learning about our topic and identifying additional search terms. In fact, automatic term mapping can save you the hassle of going to the mesh database to look up the controlled vocabulary for each of your concepts. Without much work, we were able to recognize that electronic health records and analgesic comma opioids are important mesh to include in our search. But automatic term mapping is not always accurate. Let's say instead of electronic medical records, we search the abbreviation EMR. In this case, when we run the search, it's clear that something went wrong. This article Musical Sounds, Motor Resonance and Detectable Agency is clearly off topic. Looking at the search detail shows that the automatic term mapping matched EMR to the journal Empirical Musicology Review rather than electronic health records. And sometimes automatic term mapping will map the term further than we need for our search, such as nursing. The term nursing maps to nurses and nursing, however it also maps to breastfeeding. If you're doing a search on nursing in the PICU, you will get articles like the effect of body and mind stress releasing techniques on the breastfeeding of full term babies. In cases like this one, we are getting a lot of noise due to the broad mapping. The final point about automatic term mapping is that this feature can cause problems when you are conducting a systematic or comprehensive literature review. For these types of reviews, we recommend that you take the time to manually select and format the appropriate search terms. This removes any potential errors caused by automatic term mapping and provides a stable search that others can easily reproduce. If you decide that automatic term mapping is not a feature you want to use, you can turn it off. In fact, you may be turning it off without even realizing it. For example, if I use double quotes around electronic medical records, you'll notice in the search details that PAMED didn't include the mesh. This happened because phrase searching is one of the ways to turn off automatic term mapping. Truncation and field tags other than all fields also turn off this feature. We can see that when I typed our original search electronic medical records and opioids and prescribing patterns with the TIAB field tag after each of our search terms, the search detail shows that PAMED is just searching for articles that contain these terms in the titles and abstracts. So the benefit is if you turn off automatic term mapping, you won't have to worry about PAMED mistakenly mapping your search terms to the wrong mesh or causing too much noise. Conversely, the drawback of turning off automatic term mapping is that if you want to develop an exhaustive search, the burden becomes yours to identify the proper mesh and think of synonyms to include in your search. There's one more piece of what PAMED is doing behind the scenes that's important to know about. We can see it in action whenever we conduct a search. When we type opioids and prescribing pattern and electronic medical records in the search box, without doing much else to our search, relevant articles appear at the top of the page. Automatic term mapping is helping us retrieve relevant results but the reason we see these relevant results at the top of the first page is due to the best match feature. By default, PAMED displays results using an algorithm to connect you to those results that best match your search topic. This algorithm calculates the weight of each citation based on how many terms from your search are found and the fields in which they are found. In addition to search terms, recently published articles are given more weight than older publications. The articles that have the highest weight are sent through another algorithm that will rank them by relevance. This algorithm is based on data extracted and aggregated from PAMED searches and PAMED makes revisions as new data becomes available. Because of this algorithm, PAMED results are sorted so that relevant articles appear first. As you scroll down the results, your results will be more and more off topic. Just like automatic term mapping, the best match feature is a great help when you quickly need to find a handful of articles. But there are times when it gets in the way of finding the results that you need. For example, let's take a second look at our opioid search. From the titles, it's clear that they are in line with what we're looking for. But if we look closer, you'll notice that they are from 2019 and 18. In fact, we won't find a more recent article until about halfway through the list. If you were searching a topic where finding the most recently published articles is crucial. For example, if you want to see how prescribing patterns were changing due to COVID, you will want to change how your search results are sorted. You can turn off the best match algorithm by clicking on the display options button on the right hand side of the results screen. Select publication date to see results sorted from the newest to oldest. The PAMED search box is designed to be user friendly and retrieve relevant results. This is all due to the programming behind PAMED's automatic term mapping and best match features. These are powerful tools that can save time, but make sure they are processing your search as needed. Remember to always check your search details to make sure automatic term mapping is mapping to the terms the way you want it to. Also remember, if you are looking at a topic that has had a lot of recent publications or new findings, you may want to change your display options to publication date.