Improving your search strategy: date limit filters (2/2)

Amy Cromptom (Systematic Review Analyst) and Tom Macmillan (Consultant - Systematic Review)

This blog post follows on from ‘Improving your search strategy: Randomised controlled trial filters’.

Electronic database searches can be run on multiple platforms. This article focusses primarily on applying date limits in Ovid.

 

When to use date limit filters

Study selection criteria for systematic literature reviews (SLRs) may specify publication date limits, for example:

• Only studies published after a drug or device approval date are relevant in an SLR searching for real-world evidence of clinical efficacy and safety
• Date limits may be needed in SLRs of rapidly evolving fields where previous interventions have become obsolete
• Health technology assessment agencies often require SLRs to be conducted within 6 months prior to submission ­– an SLR using date limit filters may be performed to update a previous SLR.

Various date limit filters can be used to search different electronic databases. Like other filters applied to search strings, these influence the sensitivity (number of relevant records retrieved) and precision (number of irrelevant records retrieved).

 

Limiting searches by year

In Ovid, a built-in ‘Publication Year’ filter can be used to limit a search from a specific year, for example ‘2015 to current’ (Figure 1).

 

Limiting searches using specific dates

Limiting searches by a specific date, rather than just publication year, is more complex. One option is to use the ‘Date of Publication’ field. However, citation formatting is inconsistent in Embase (some publication dates are structured DD MMM YYYY [e.g. 18 JAN 2018]; others are unstructured [e.g. Jan-Feb 2018, or Winter 2018]) (1, 2). In Medline, the ‘Date of Publication’ is consistently formatted (YYYY MM DD); however, the month and/or day are not present in every record (3). This variability makes ‘Date of Publication’ unsuitable for limiting a search by specific dates (1). Therefore, other date limiting fields must be selected and these vary across databases.

Date limit filters available in Embase

Several date limits are available when searching Embase in Ovid, all formatted YYYYMMDD (2):

• Date Delivered (.dd.): the date a citation XML file was produced for distribution to Ovid with the state=”new”. This includes only new records; The Date Delivered is removed when a record is revised (1)
• Revised Date (.rd.): the date the citation XML file was produced for distribution to Ovid with the state=”update”. This date can change if an updated record is delivered to Ovid
• Date Created (.dc.): the date of final activity on a citation before the XML file is produced for distribution to Ovid. This date is always present in an Embase record. The Date Created is not removed when a record is revised but will change as a new record is added. There may be a short lag between the Date Created and Date Delivered while the record is processed and added to the database (1)
• Entry Week (.em.): the year and week (01–52) when records were added to Embase on Ovid, since the yearly database reload. The Entry Week includes all records but can only be used within the current year (1). ­­

Date limit filters available in MEDLINE

Several date limits are available when searching MEDLINE in Ovid, all formatted YYYYMMDD (3):

• Create Date (.dt.): the date a record is added to PubMed before processing begins (adding Medical Subject Headings [MeSH] terms, completing quality validations, and distributing the record to PubMed). This is not the same as the publication date. This date remains the same throughout any revisions
• Revised Date (.rd.): the date a record is revised or corrected in Ovid MEDLINE
• Entry Date (.ed.): the NLM internal Date Completed, when processing of the record ends. The entry date may change when a record is revised, and some records do not have an entry date
• Update Date (.up.): the date a record was added to MEDLINE since the yearly reload completion. This date changes with each reload reflect the date Ovid starts processing the reload data, so can be used as a limit only within the current year
• Electronic Date of Publication (.ep.): the date a record was sent to NLM for inclusion in MEDLINE
• Entrez Date (.ez.): the date a citation was added to PubMed
• MeSH Date (.da.): the date MeSH terms were added to a citation. The MeSH date is set equal to the Entrez date until MeSH terms are added.

Date limit filters available in Cochrane databases

When searching Cochrane Central Register of Controlled Trials in Ovid, the only available date limit is Ovid’s built-in ‘Publication Year’ filter. When searching Cochrane Database of Systematic Reviews in Ovid, three date limit fields are available (4):

• Date of Publication (.dp.): the month, year, and possibly day that the journal was published. This can only be limited by year
• Date of Most Recent Amendment (.dr.): the month, year, and possibly day that the review or protocol was last updated. Searches can be limited by a range of years, or searches can be limited to more specific dates, including day and month, only for a single date, not a date range (format, MM-DD-YYYY)
• Date of Most Recent Substantive Amendment (.ds.): the month, year, and possibly day that the review or protocol was revised in a substantial way. Searches can be limited by a range of years, or searches can be limited to more specific dates, including day and month, only for a single date, not a date range (format, MM-DD-YYYY).

 

Working example

In a recent project, SLR search strings were limited to identify evidence published after the 1^st January 2010. As part of the search string development process, a set of relevant records were identified to validate the search strategy. In the first iteration, Date Delivered (.dd) was used when searching Embase in Ovid, missing 13 out of 37 relevant records. To improve the sensitivity of the search string, we assessed two alternative date limit filters:

• Publication year
• Date Created.

Both filters identified all 37 relevant records. As there was no requirement to limit the search to a specific date or month within a year, the best approach was to use the publication year filter. However, if there had been a requirement to limit the search to a specific date within 2010, the Date Created filter would have been the most appropriate to identify all relevant evidence.

 

Conclusions

Each date limit filter varies in sensitivity and precision. Although the Date Delivered field in Embase is commonly used to limit searches, our experience is it can reduce sensitivity. We have identified alternative methods that use other fields, increasing the search sensitivity. Combining the Date Delivered and Revised Date limits in Embase, or the Create Date and Revised Date limits in MEDLINE, retrieves the most studies. This increased sensitivity reduces the risk of missing relevant studies. However, the precision is likely to be reduced, resulting in retrieval of more irrelevant studies, as records revised within the date limits but published earlier may be included. Therefore, the date limit filter, or combination of filters, should be chosen on a project-by-project basis.

If you would like to learn more about systematic literature reviews, please contact Source Health Economics, an independent consultancy specialising in evidence generation, health economics, and communication.

 

References

1. Shete S. How can I limit search results by a date range in Ovid? : Wolters Kluwer Health; 2020 [Available from: https://wkhealth.force.com/ovidsupport/s/article/Limit-by-date-range-in-Ovid.
2. Wolters Kluwer Health. Embase: Excerpta Medica Database Guide 2021 [Available from: https://ospguides.ovid.com/OSPguides/embase.htm#em.
3. Wolters Kluwer Health. MEDLINE® 2020 Database Guide 2020 [Available from: https://ospguides.ovid.com/OSPguides/medline.htm#ED.
4. Wolters Kluwer Health. Evidence Based Medicine Reviews: Cochrane Database of Systematic Reviews 2014 [Available from: http://ospguides.ovid.com/OSPguides/cochdb.htm.