EMERSE User Guide

A Temporary Patient List is a list that a user can make and use during a single user session but will not be saved between sessions unless it is converted to a Saved Patient List. That is, it can be used for a single session but is erased when the user logs out. Such a list is useful in cases where a patient, or set of patients, needs to be reviewed but there is no need to save the list of patients for another time. A "review preparatory to research" should utilize this Temporary Patient List since such a list should not be saved during this type of research review activity.

A Temporary Patient List is meant to be fast and simple, and has a limited set of features. By contrast, a Saved Patient List (see next section) provides more features. For example, with a Temporary Patient List it is not possible to share the list with other users, nor is it possible to annotate the list with Comments or Tags, both of which are only possible with a Saved Patient List.

A Saved Patient List is a list that a user can make and save between sessions. There is no limit to the number of Saved Patient Lists that a user can have, and each list can contain between 1 and 100,000 patients. Because these lists are saved, each list should be given a name and a description, available under Name/Description. A Saved Patient List is very similar to a Temporary Patient List but there are several notable advantages when using the Saved Patient List.

Patients identified using the Find Patients feature can be moved to a Temporary Patient List for more detailed searching and viewing of their notes. Once the Find Patients search is completed, simply click on the button called Move to Temporary Patient List. This will transfer the set of patients identified from a Find Patients search to a Temporary Patient List, which is not saved between sessions. Once that move occurs, the list of patients will be shown. To then change that Temporary Patient List to a Saved Patient List (the latter of which is saved between sessions), simply click on the option called Save.

In addition to moving patients to a new Temporary Patient List from a Find Patients search, patients can also be added to a list (either to an existing list or to a new list). This process is the same regardless of whether you are using a Temporary Patient List or a Saved Patient List. Entering patients is done via the Add Patients option for the patient lists. To enter new patients, you can type each MRN one-by-one into the MRN text entry box, or you can paste an entire set of MRNs at once. For example, you can copy and paste an entire column of MRNs from an Excel spreadsheet and paste them into the MRN text entry box to create a patient list. If there are already patients in a patient list, adding new MRNs will append them to the end of the existing list, up to the maximum number allowed.

Any time an MRN is entered, that MRN is checked against the complete set of patients within EMERSE. If an MRN is not found, a warning to the user will be provided. Similarly, if a duplicate MRN is entered for a list, the duplicates will be removed automatically and the user will be alerted.

As described in the section on Temporary Patient Lists, Encounter IDs (EIDs) can also be entered into the Add Patients section. Each EID is associated with an MRN, so if you start with a list of EIDs, you can use that EID list to generate your patient list based on the corresponding MRNs in the system. This will only work if your local EMERSE instance has EIDs loaded into the system.

The Compare Patients Lists feature lets you compare two Saved Patient Lists to each other with respect to the patients contained within each list. This feature can be accessed via the Compare Patient Lists button which can be found above the table displaying the Saved Patient Lists.

Compare Patients Lists provides functionality that can be useful for separating out only those patients of interest that may be found among two lists. This feature essentially builds a Venn diagram showing the overlap of patients between any two Saved Patient Lists. You can then select any part of the diagram you are interested in simply by clicking on the appropriate section of the diagram. Once your selection is finalized, click on the Create New Patient List From Selection button and you will then have a new Saved Patient List based on your selection (you will have to provide a Name and Description for the new list). Note that when lists are merged, duplicates will be removed automatically. The original lists used for the comparison will not be modified.

It may be useful to use the Compare Patient Lists feature when you want to compare patients found through different approaches for obtaining the patient lists. For example, suppose you find a set of patients who had "back pain" and a separate set of patients who had "acupuncture". Using the Compare Patient Lists feature you can find the set of all patients who had both concepts mentioned in their notes at least once, even if the mentions were in completely separate notes or sections of the EHR.

Or, suppose you have a set of patients with a diagnosis of "amyotrophic lateral sclerosis" (List 1) and now you want to exclude any patients with a mention of "packs per year" (List 2). You can make two separate lists, one based on each search. Then using the Compare Patients Lists feature, you can select out the list of patients from List 1 that did not also appear in List 2.

Finally, some have used this feature to reduce a list obtained from an external source to make it more relevant. For example, suppose you have a list of 10,000 patients from a registry and now you want to know which ones mention a certain drug, paclitaxel. You can upload the list of patients from the registry (List 1). Separately you can run a Find Patients search for the word paclitaxel (List 2) and then use the Compare Patient Lists feature to identify only those patients that appeared in both lists. As of EMERSE version 6.1, this same kind of result can also be achieved by using the Find Patients feature applied to the original patient list, followed by Tagging the patients of interest and then clearing the Untagged patients.

Terms should be added (and edited) in the section called Edit, specifically into the small text box where it says "Enter Terms/Phrases (one at a time)". These can be single words (chemotherapy) or multi-word phrases (acute lymphoblastic leukemia). One term or phrase should be entered at a time. This is so that the system can help provide suggestions about what has been entered and to help with selecting the desired colors for each term. Once a term has been added, it shows up in the right column of "Active Terms/Phrases".

As mentioned in the section on NLP, some terms within EMERSE have been pre-identified and mapped to UMLS Concepts. These concepts are codes (Concept Unique Identifiers, or CUIs) developed by the National Library of Medicine (NLM) that can, in turn, be used to map back to other standard ontologies and vocabularies. The official site to look up concepts, is the UMLS Metathesaurus Browser.

Concepts are always in the form of the letter C followed by 7 digits. For example the term chest pain has the CUI C0008031 and is mapped to R07.9 (ICD-10-CM), 786.50 (ICD-9-CM), HP:0100749 (HPO), and more. It is important to note that the concepts are being used 'behing the scenes' in EMERSE to search for words and phrases that have been labeled with a CUI. The actual CUIs themselves are not being searched. That also means that when the results are displayed, the terms and phrases will be shown and highlighted, and not the CUI itself.

You can search by UMLS concept simply by entering a CUI into the search term box. EMERSE will recognize it as a concept and show it in a slightly different (monospace) font, with a small light bulb icon next to it to help identify it as a concept. It is even possible to mix regular terms with concepts. For example, you can combined the word severe with the concept for migraine (C0149931) to look for that specific phrase. The figure below shows a search for C1690547 (nickel allergy) as well as severe C0149931 (severe migraine).

There are multiple options for customizing how terms should be searched. This ranges from whether the term should be search in a case-sensitive manner to whether it should be search in a negated, uncertain, or other manner. Each of these options are described below. All options, other than the color, have a "subway icon" next to them to help provide a high-level visual summary of the selected options on other parts of the screen (where it shows the "Active Terms/Phrases"), which will be described later. These icons are inspired by the "service bullets" used by the New York City subway system.

Below is a screenshot of the term entry screen, followed by a description of each option.

If you have a large list of terms that you want to enter, you can paste in an entire list of terms at once through the Bulk Upload feature. Simply click on the Bulk Upload option, and then type or paste in the list of terms into the text entry box, one term per line. You can also choose settings for the terms (such as excluding Negated terms, etc), but the settings (and the term color) will apply to all of the terms entered.

Note that some of the options for entering terms individually will not be avialalbe, including the Proximity and the Fuzzy Match

Terms already entered will appear in a column to the right of the panel where terms are originally entered, with a header labeled "Active Terms/Phrases".

Editing terms can be done within the Edit section, which is also the same place where you can enter new terms. Terms can be edited if you are the owner of the Saved Terms list (for Temporary Terms there is no concept of ownership since it is only a temporary list that you create yourself). If you are not the owner of the Saved Terms you will not be able to make any changes. To edit an existing term, just click on the pencil icon inside of its "bubble" and an editing box should appear. This is also how you would change the color of a term, or any of the other settings for that term.

Do not forget to click the Done button to save the changes/edits.

There are several ways to remove terms from a Saved Terms list. If you want to remove a single term, you can click on the pencil icon for a term and then click on the Remove button.

You can also "grab" the term in the upper right corner of its colored "pill" and drag it out of the area. You may see it turn into a trash can icon and then it will be gone when you stop dragging.

Finally, you can delete multiple terms quickly by holding down the X key on your keyboard (make sure that the focus is not on any term entry box). At that point you should see small trash van icons on all of the term "pills" and then while holding down the X you can click on each trash can icon and those terms will disappear.

Below is a short video showing these various options.

If you want to remove all terms at once from a Saved Terms list (aka Term Bundle), click on the Clear/Delete tab and then click on the Clear All Terms button. This will remove all of the terms but preserve any sharing preferences that you may have used with the Saved Terms. To delete a Saved Terms list entirely, choose the Delete Term Bundle button. This will completely remove the Term Bundle, but be careful because there is no undo option once it has been deleted.

EMERSE will check the spelling of terms entered. While typing the term spelling suggestions will displayed under the Terms text entry box in bold, preceded by "Did you mean". For example, if you enter your terms as word misspeling you will see:

Did you mean "word misspelling"?

To correct the spelling you can simply click on the spelling suggestion and the misspelled words will be replaced with the corrected version. Of course, sometimes you may want to search for misspellings, in which case just leave those spelling errors as-is.

For some terms you type in, you might see an Estimated Document Count appear. This is available to provide a high level estimate about how many documents (not patients) contain that term. These counts are based off of any Synonyms loaded into the system (Synonyms are described later in this guide). The estimated counts are not real-time counts (real-time counts have to be done through searching), nor are they adjusted for any types of filters or other criteria. Synonyms data may be periodically updated by your site administrators and the estimated counts are updated at those times. It is worth nothing that if you turn on the counts display with the Synonyms option, these are the same counts, but shown in a different location on the interface.

Synonyms are a very powerful feature of EMERSE, and are described in detail in a separate section.

A set of Temporary Terms can be saved which will allow you to keep and re-use the set of search terms permanently and share it with other users. Simply cluck on the Save option. You will be asked to provide a Name and Description and then that list of terms will be saved as a Favorite within your list of _Saved Terms, which are described in detail below.

Saved Terms are grouped into two categories: Favorites and All Available. The Favorites is a smaller subset of all potential Saved Terms (also called Term Bundles) that are available to you, so that you can have a smaller list to work with. All Term Bundles that you make yourself will automatically be in your Favorites. Otherwise, to make a Term Bundles that is not yours into a Favorite simply click on the Favorite checkbox.

Using a Term Bundle is simple: just click on the row in the table with the Term Bundle of interest, and EMERSE will display the contents of the Term Bundle. Then, just click on either Highlight Documents or Find Patients.

Making a new Term Bundle is also very easy—​just click on the Create New Saved Terms List button and give the Term Bundle a Name and a Description. A few other features of a Term Bundle are described below.

All Term Bundles have a concept of "ownership". That means that only the "owner" (the person who created the Term Bundle) can manage it, such as adding/editing terms and choosing how the Term Bundle should be shared. Others can use it with the right Sharing privileges but they cannot modify it.

If you have trouble finding the right Term Bundle, you can try to use the Search feature, which can be found in a small text box above the list of Saved Terms. Just start typing what you are looking for and those matching what you type will remain and everything else will be hidden. Search looks across the Name, Description and terms within the Term Bundle. You can not search for the Owner of a bundle, but you can sort the list of all available Saved Terms/Term Bundles based on owner.

If you are not the owner of a Term Bundle but wish to make edits, you can make your own copy and become the "owner" of this new Term Bundle. To access this feature, click on the Copy tab of the Bundle. This new Term Bundle will then be yours to modify and control, but any changes to the original Term Bundle will no longer be reflected in your new Term Bundle, as they have essentially been split off from one another. You can also make a copy of your own Bundle if you want to make some changes but leave the original one intact.

Similar to Patient Lists, Term Bundles (also called Saved Terms) can be shared with other users. As of version 6.4, this can include users at another EMERSE site, as long as that site participates in the network. To access this Sharing feature click on the Share tab. There are four options for sharing these Bundles, described below:

Boolean searches use the operators AND, OR, and NOT. They must be capitalized or they will be considered as regular terms; that is, it will search for the words and, or, and not if they are lower-case. NOT in this context means that a document should not be identified/returned if it contains that term. For example:

The query above would identify documents that have the words chest and xray in the same document (anywhere in the document), but if the word abdomen was also in the document it would not be considered a hit (i.e., would not be returned) even if the first two words (chest and xray) were there.

You can also group booleans together using parentheses, for example:

This query will return documents that have the terms chest, pneumonia and headache all together in the same document in any order, or documents that have kidneys, reflux, and back pain all together in the same document in any order.

Two basic types of wildcards are supported (? and *). Wildcards can only be applied to single words, and cannot be applied to phrases.

? is used to match a single character. For example:

Would match words like beam, beat, bear, bean, bead, etc.

* is used to match zero or more characters:

Would match terms like hypertension and hypertensive, as well as just hyperten.

A fuzzy search allows you to search for terms that are similar to the term of interest. This can be way to look for possible misspellings. A fuzzy search can only be applied to a single word at a time. To use a fuzzy search, append the term of interest with a tilde (~) followed by a number between 0 and 2. These numbers represent the number of potential character substitutions that would have to occur to match the new term.

For example:

Would match beats as well as the word breast, since both of those terms can be created from the word beast by changing just one character. (Note that the original term beast would also be highlighted in this example).

A proximity search looks for two words within a certain maximum distance from one another, regardless of their order. Separate the two words by a space and wrap them in double quotes, then add a tilde (~) followed by a number representing the maximum distance in words that they can be separated.

For example:

Would match phrases like …​a CT scan showed a mass in his left…​

as well as …​a 12x12x8 cm mass in her L upper lung found on CT…​.

and even simply mass CT

A Regular Expression is powerful way to match text using complex patterns. Regular Expressions are commonly used by computer programmers, and it is expected that only those with a certain level of technical expertise would ever use this feature. The syntax for creating Regular Expressions can be complex and difficult to understand, and can vary between programming languages. To create a Regular Expression with EMERSE follow the guidelines for Java regular expressions.

Regular Expressions within EMERSE can be created for single words/tokens only (not multi-word phrases). Nevertheless, there may not be much advantage to using this featured compared to simple wildcard matching. Wrap the regular expression within a forward slash at the beginning and end to tell the query parse that it is a regular expression. Also, it is necessary to use one of the two "hidden" prefixes for the terms (tx_ for case-insensitive or TX_ for case-sensitive). For example,

Would match 100mg, 150mg, 400Mg, 650MG, etc.

With the exception of regular expressions, searching the index for case-sensitive terms is no longer possible with Advanced Terms as of version 7.x.

Metadata are additional pieces of data about a document that can potentially be used to help with searching. Examples of metadata might be the source of the document (e.g. the main EHR, radiology, pathology, etc), the type of test (e.g., MRI report, CT report, plain film report, etc), the clinical service (e.g., nephrology, cardiology, rheumatology, etc) and others. While it is possible to search based on metadata, the specific types of metadata and names of the possible elements will vary from one installation/site to another and will depend completely on how the system was configured locally, what data have been loaded, etc. Therefore, if you are interested in this option you will likely need to contact your local EMERSE administrators to get more information. Further, the Filters in EMERSE are applied to the Advanced Terms searches, so it is preferable to use the standard Filters rather than try to recreate the complexity of a filter in raw Solr queries.

Below is an example of the type of query that could potentially be run if specific metadata elements were captured and stored within EMERSE:

The query above will search for the phrase motor vehicle accident and it will limit the search to a locally named source that is called EHR and a locally named department called ADULT EMERGENCY. As mentioned above, the specific names of sources, departments, and other metdata will be unique to each installation, so it is not possible to provide specific details on what options will be available at each site.

Searches can be limited by date ranges using the Advanced Terms feature. Currently, due to technical constraints for how EMERSE was designed, this date feature can only be used when searching across All Local Patients using the Find Patients function. It will not work when searching across a Patient List with the Highlight Documents function. At some point the system may be re-designed to allow these Advanced Terms date ranges to be used everywhere.

It is worth pointing out that there may be no real need to use a date range within an Advanced Search query, since a date range can already be applied to any search (even searches using Advanced Terms) by using the Filters feature. However, if using dates in the query is desirable, they should be in the standard Solr date syntax, which can be somewhat complex to use. It is also worth pointing out that if you set a specific date range using the Filters AND you also set a custom date in your Advanced Terms query, an error will occur and 0 patients will be returned. In other words, if you want to use the date options within an Advanced Terms query, you need to leave the standard EMERSE date fields blank.

A basic example of how a date range can be included in a search is below. This query will search for the term trabeculotomy and it will limit the search to the date range January 20, 2015 through May 20, 2015.

The following query is similar to the one above, but shows how some basic date math can be used. In this query the same date is used for both the initial and final dates (2015-01-20), but the second instance of the date is followed by +4MONTHS which essentially adds 4 months to the date, and thus is equivalent to the previously shown query that includes the date range of January 20, 2015 through May 20, 2015.

A few more examples, using the special NOW option are below.

Search for trabeculotomy between January 20, 2015 and the current date:

Search for trabeculotomy between 12 months prior to the current date and the current date:

Search for trabeculotomy between 30 days prior to the current date and the current date:

Search for trabeculotomy between 60 days prior to the current date and 30 days prior to the current date:

EMERSE supports the ability to incorporate multiple Synonym Datasets. As mentioned above, these datasets can only be added by a system administrator, but once they are added they can be available to all users. The Synonym Preferences page allows users to have granular control over these Synonyms including the ability to include or exclude them from the suggestions.

There are several options available to users from the Synonym Preferences screen:

Enabled: When this is on, the synonyms will be available to the user and will be included in any suggestions if matches are found.

Counts available: If counts are available, they will be shown in parentheses next to suggested the term itself. Counts represent how many documents these terms appear in across the entire corpus of documents that EMERSE has indexed. These counts can be useful to help determine which terms might be most valuable in the search (i.e., yield the most results). For example, terms with counts of 0 are not in the full index of all documents and won’t yield any results, so they do not need to be included. Note that counts are not updated daily, so if new documents are added the counts will not change unless the EMERSE administrators starts the counting process again (which can take days to run). Additionally, if counts are available they can be sorted by frequency when the Synonyms suggestions are shown.

Show Counts: Turning this on will show the counts if they are avaiable, and turning it off will suppress the display of counts.

Show Zero Counts: If counts are available, then keeping this option off will prevent the system from suggesting terms that have a count of 0. There is really no reason to show terms with counts of 0 other than to see the wide range of potential terms available. Turning this on will show all terms, even if they have a count of 0.

When in the Find Patients mode, EMERSE will provide the following results in different sections:

If you are satisfied with the search results, you can then explore the list of patients in more detail, or save the list for later use by clicking on the button Move to Temporary Patient List. Note that to actually save the list for use in future sessions you would then need to click on the Convert to Saved Patient List button after converting it to a Temporary Patient List. Once the patients have been moved to a Temporary Patient List, you can also search through their documents using the Highlight Documents button, because at this point you have switched from the Find Patients mode to the Highlight Documents mode by selecting a subset of patients in the system to look through.

If you used Find Patients against a Patient List you will also see an option to Tag Patients in List. This will add a Tag to each patient that was in the result from the current search. More details can be found in the section on Tags.

With Find Patients, a patient will be included in the results if any of their documents match the search criteria. The way the search runs is different depending on whether it is run against All Local Patients in the system, or against an existing Patient List.

For example, if you search for heart and "chest pain" across All Local Patients, then a document must contain both the word "heart" and the phrase "chest pain" to match, and a patient must have at least one document that matches those terms since the terms are different colors. If a patient merely has a document containing the word "heart" and another different document containing the phrase "chest pain", the patient would not match because the requirement is that the terms appear together in the same document.

If the same search for heart and "chest pain" os done within a Patient List, then a document matches if it contains either the word "heart" or the phrase "chest pain"; the document does not have to contain both. For example, a patient will match the query when they have a document that contains the word "heart" even if they have no document that contains the phrase "chest pain".

Document matches for Advanced Terms operate a bit differently: the results are based only on the search syntax used, not the highlight color; all advanced searches are highlighted in yellow. Terms in advanced search not separated by a boolean operator are searched as if they were separated by the OR operator. Parentheses should be used to make grouping clear.

Advanced term search can do things that have no analogy in Temporary Terms or Saved Terms, such as (heart AND "chest pain") OR "chest tightness" which matches documents that either contain both "heart" AND "chest pain" OR just contains the phrase "chest tightness" (this is generally known as 'nested Booleans'). In this example, a document containing the phrase "chest tightness" is enough to match, but a document containing the phrase "chest pain" without the word "heart" is not enough to make a match.

Because of how searches are conducted, and how patients are identified when a search using Find Patients is applied to All Local Patients, it is best to ensure that true synonyms for a concept are highlighted in the same color. Additionally, be careful about adding too many terms especially if a term is a different color from the others, since it means that the additional term must also be in the document for the document to be retrieved. Adding an extraneous term that isn’t very important for identifying the right patient population has the potential to result in no patients being found. For example, if the following query is used with a search across All Local Patients, the query will only find patients that have at least one document with all three of those terms in the same document:

Thus, it is better to use only those terms that are absolutely necessary to find the right patients, or to use a Term Bundle and set the color of the terms more strategically, such as:

In this case, the above query will look for documents that contain heart AND either chest pain OR chest tightness, but chest pain and chest tightness do not have to both be present, only one of them.

When in the Highlight Documents mode, EMERSE will provide results including:

The Highlight Documents mode is simpler than the Find Patients mode. This is because all terms are treated as being separated by the OR operator, regardless of the color in which the terms are highlighted (unless otherwise specified using actual Boolean operators with the Advanced Terms). The general idea is that since the patients of interest have already been identified, the goal with Highlight Documents is to simply highlight any terms of interest anywhere in the documents to support a rapid, efficient chart review process. As a result, there is no harm (other than speed/performance) in including additional terms even if they are not present anywhere in the documents, since their absence will not affect the highlighting of the other terms.

In general, no specific search strategies are needed when searching in the Highlight Documents mode. Even so, it may still be advantageous to group terms with similar meaning by the same color to aid in more rapid chart reviews. In other words, if you are searching for a list of narcotic medications and the resulting constipation that the medications might cause, it may be useful to set all of the narcotic medications to be green, all of the constipation terms to be orange, and all of the stool softeners/laxatives to be purple. Then, someone reviewing the charts would only have to recognize one of three colors to know what kind of concept was being highlighted.

This is described elsewhere, but it is extremely important to understand the differences between the two search modes, Find Patients and Highlight Documents. The way EMERSE works differs depending on which search mode is being used, and therefore this difference must be understood to ensure effective search results.

EMERSE can provide a lot of suggestions for additional terms through the Synoyms feature. Use this to expand your search to cover more possibilities for how something was worded or phrased in the notes. Also be careful because EMERSE is a very literal search engine. If you want to look for antibiotics and that is what you type in, the system will look for the word antibiotics only. If what you really meant was to find any instance of any antibiotic mentioned in the notes, then you will have to include all of the possible terms for these, and there may be hundreds of them (e.g., cephalexin, Keflex, Ancef, Ciprodex, amoxicillin, etc). The Synonyms feature can certainly help expand the breadth of the query terms, but it may not be complete.

While the Synonyms can be very useful and powerful, be careful about what Synonym suggestions you add. EMERSE has the potential to provide a very large list of options, but adding too many terms could slow things down without providing much benefit. And, for ambiguous terms, you might inadvertently include concepts you don’t even want. For example, if you are searching for synonyms of MI because you want to look for terms related to myocardial infarction, EMERSE will also suggest terms like Michigan since that is also a different meaning of MI.

Similar to trying multiple variations, be careful about how specific the query is. For example, if you enter "metastatic breast cancer", then the system will look for that specific phrase. It might be better to look for the phrase metastatic and the phrase "breast cancer" separately, so that you have a greater chance of finding other variations such as:

Note, however, that sometimes breaking up the query can lead to false positives if, for example there was text such as:

Another approach for 'separating' the terms is to use the Proximity Search feature. This will allow you to search for two terms within a specified number of words from each other. For example, suppose you were looking for "sublingual nitroglycerin". A logical alternative phrase would also be "nitroglycerin sublingually". But if the phrase in the clinical note is actually "nitroglycerin was given sublingually" then the simpler, shorter phrase would miss it. Including the terms in a Proximity Search would capture the longer phrase. A Proximity Search within Temporary Terms or Saved Terms will search up to 5 words apart. To search with even more space between words, the Proximity Search can be used with the Advanced Terms. Just remember that the more spacing between two words, the more likely that there is no true connection to them which can result in a false positive result.

A concept that you’re looking for may be implicitly described but not explicitly stated. An example of this would be looking for the concept of metastatic as it relates to breast cancer. If the phrase is…​

…​this implies that the cancer was metastatic even though the term metastatic or its variants (metastasized, mets, etc.) was not explicitly mentioned. EMERSE, being a simple tool, is not able to interpret that phrase as being an example of a metastatic breast cancer.

Another example:

This is an example of an asymptomatic rash, even though neither of those two words (neither asymptomatic nor rash) appears in the sentence above. This can happen often, so it is important to be careful of what might be missed when only searching for specific phrases or concepts.

There are hundreds of terms that are used in the clinical notes that are ambiguous—​that is, it is hard to know what a term represents without understanding the context in which it appears, and some terms/abbreviations may have been completely made up locally by a single clinician or a small group of clinicians.

For example, diffs might mean differentials, or differences, or difficulties or even be a part of an abbreviation for Clostridium difficiles (C diffs).

Other common ambiguous abbreviations include CA which can mean cancer, calcium, or California, and MG could mean milligrams, myasthenia gravis, or magnesium. There are many more examples like this.

Even terms that one might think are straightforward might not be. For example, imagine you are interested in exploring milk consumption to assess dairy intake, so you search for milk. Be careful because there can be terms like almond milk, carob milk, soy milk, rice milk, and others which are clearly not the right kind of milk. This is why reviewing the results carefully is very important. Also note that you could add those 'false positive' terms as an Exclude Phrase to remove them from the search results (see section on: E Exclude Phrase).

You may be tempted to use the most standard, common phrasing for a particular concept, but it is best to consider all of the ways in which something might be phrased. Clinicians will sometimes even make up new terms. For example, scaphotrapeziotrapezoidal may be the official spelling, but clinicians have also used terms such as scaphotrapeziumtrapezoid, scapho-trapezoid-trapezium, and scaphoid-trapezium-trapezoidal. Clinical note creators also add hyphens or spaces (these are equivalent in EMERSE) into words where they might not be expected, as can be seen in the above examples, or even simple words like ham burger or yester-day. The system-suggested Synonyms can help capture some of this variability lot, but they may not be enough to cover all of the possibilities for what is mentioned in the notes.

Remember that the people who create clinical notes often have a difficult time spelling medical words, or they make errors when they type, or automated transcription systems can "hear" the wrong word, etc. For example, beast mass and prostrate cancer are phrases that contain correctly spelled words, but they are not the right words. In fact, many such word substitutions of correctly spelled (but incorrectly used) words appear. Often these are homonyms, but not always. Think carefully about how someone (or a computer) might hear a word and insert the wrong word. If you’re looking for people with a hoarse voice, be sure to search for horse voice because it’s almost guaranteed that someone will have used the wrong word at least once.

Following is a table containing examples of such word substitutions that have been found int he medical record. It is by no means complete.

This has already been described in the section on Terms, but it is worth pointing out again. EMERSE ignores multiple characters such as ! ? . > < + - #. This means that some searches will not work. For example: BRCA1- will be interpreted by EMERSE as simply BRCA1, which means that although you can still search for BRCA1 you cannot distinguish between BRCA1- and BRCA1+. (If the phrase is "BRCA1 negative", or a similar variation, then EMERSE will be able to find the concept. Similarly, "150 pounds" when written as "150 #" or 150 # will simply be interpreted as 150 and a blood pressure written in the form of 120/90 will be interpreted the same as 120 90, meaning that you cannot use the slash symbol (/) as a way to look for a pattern that appears to be a blood pressure.

A search engine like EMERSE is great for finding concepts quickly, but it does not have any understanding of what it found, or the context in which a term was found. That is why it is important to carefully review the results for accuracy. Even in cases where you might think no error is possible, it just might be possible. The following is a real example of what was found in a clinical note:

Similarly, terms like ular could represent a misspelling of the word ulnar or part of a word that had an inadvertent space such as ventric ular or fib ular.

Another example:

As a general rule of thumb, if you have found what you were looking for, and have verified that it is truly the correct concept then consider that to be a success. If you haven’t found what you were looking for, the concept may not be there at all, or it may be that you just haven’t figure out the unusual, or incorrect, way in which a term or concept was phrased.

If possible, talk to someone who has clinical expertise in the area for which you are conducting your searches. They might be able to provide guidance about how certain concepts are likely to be phrased in the clinical notes.

Queries that are too long or specific will generally not work well. For example, if you are looking for a positive depression screen based on a PHQ-9 test, do not search for "positive depression screen based on a PHQ-9 test". Instead it might be better to search for PHQ-9 since that alone will help narrow down the results because it is already very specific. This could perhaps be combined with terms such as positive to help narrow down the results. Or, consider a proximity search where PHQ-9 and positive are searched within 3 words of each other to help narrow things down further.

With more general terms you may need to add more terms to help narrow the results. For example, if you are looking for type 1 diabetes and you just search for the word diabetes you will end up with far too many irrelevant hits because you will most likely be finding diabetes related to type 2 diabetes since it is much more common.

In other words, you will need to come up with a strategy for how to construct your search queries with search terms, and this will usually be very customized to your specific task.

Indeed, there may be times when using longer and more specific phrases can be useful. An example of this might be to look for wording used in a specific note template or report template, which can be useful when trying to identify specific notes needed for data abstraction. In such a situation, the specific text surround a concept of interest may be needed to narrow down the results to the desired one. The example below identifies very specific language used in a test result reported by the Mayo Clinic Medical Laboratories. Searching for this very specific language, which does not appear to vary between reports, helps to identify the actual lab report and not the mention of HIV-1 elsewhere in the notes.

Sometimes it may not be an effective strategy to find your patients of interest with a single query. It may require several queries. For example, you may want to first identify your patient population using the Find Patients search mode just by searching for a rare disease. Once you have identified those potential patients, you can move them to a Patient List and then search them for more details (for medications, side effects, etc.) in a subsequent query using the Highlight Documents search mode.

Another example: Imagine you want to find a set of patients who have fibromyalgia and back pain but your query with the Find Patients mode returns almost none. This may be because with Find Patients the system is looking for documents that have the terms fibromyalgia and back pain in the same document. But what if the patients have these terms mentioned in different documents? One solution, among several, would be to first conduct a search for fibromyalgia using Find Patients and then saving the result as a Saved Patient List. Then conduct another, separate search for back pain using Find Patients, and saving this result as a Saved Patient List. Then, use the Compare Patient Lists feature and select the patients that belong to both lists. The resulting set of patients would be ones that contain the terms fibromyalgia and back pain, but not necessarily within the same document. This smaller subset of patients could then be saved as a new list and searched further using the Highlight Documents feature.

EMERSE was designed to work with unstructured, free text data. These clinical documents can be extremely valuable for finding the needed information, much of which might not be present anywhere else in the medical record. However, there are times when using structured data makes sense. For example, if you want to find all patients with a hemoglobin less than 6, you can’t type that into EMERSE and expect to get any sensibile results. It would be better to use a separate system designed for structured data (e.g., DataDirect at the University of Michigan, or the widely used i2b2 Workbench). It may also make sense to use structured data such as ICD-10 billing codes as a first pass attempt to narrow down a patient cohort, and then search that subset of patients within EMERSE to do additional verifications within the clinical notes.

A patient list generated from a different system can be imported into EMERSE, or one can simply copy and paste a column of medical record numbers from an Excel sheet containing a list of patients generated in a different system. EMERSE also provides the technical capabilities for importing a list directly from another software system, but this functionality will depend on how it has been locally installed, configured, and connected to other systems.

It is best to think of EMERSE as being one tool among many in a toolbox of useful software to support your work. It is unlikely that any single tool will enable you to complete your task, but using multiple tools together will help you efficiently get your work done.

While this may not often be possible, if you have any control over how data are initially entered into the electronic health record, you can use strategies to help identify the note or concept later when searching. One example is to use standardized phrases that can be used in search later on that will positively identify only the notes of interest. Also, using more standardized language helps with ensuring that no unusual variations in wording are missed with the search.

Give thought to how you might want to identify and abstract data later on. It may be useful, for example, to include a unique "note identifier" in a note template, so that searching for just that identifier will identify all of the notes of interest. You can use something distinctive like stntv25; it can even be simple phrases like Strep throat note template version 2.5, as long as it is unique to that set of notes.

Sometimes it helps to think 'outside the box' when conducting searches. For example, you may be interested in identifying patients based on structured ICD-10 codes. If your clinical notes contain such codes within the text of the document, they can be searched just like searching for any other word or phrase. If you’re interested in finding Ehlers-Danlos syndrome, which has an ICD-10 code of Q79.6, just search for that term and it will find any mention of that code in the notes. Of course, since the dot (.) is ignored, you could simply search for Q79 6 to get the same result). The main thing that you cannot do, as already mentioned above, is search for numeric data using comparitors such as greater than (>), less than (<) or equal to (=).

Or your access has been revoked. EMERSE contains patient data, so it is not the kind of system you can just log onto without someone providing an account. Talk to your system administrators if you think that is the issue.

If you can’t even see the login page (because the page is not found), this may be the reason you’re having trouble. In general, EMERSE will be available on a secured network available only to individuals on that internal, trusted network. There may be ways to get to this network from the outside using virtual private network (VPN) software. This is an issue to bring up to your local IT support teams.