Information retrieval in uncertain times

Alternatives to the US databases of the National Library of Medicine (NLM)

Summary

• In Germany, information retrieval relies heavily on freely accessible medical information services from the United States. Currently, there are no comparable German, European or international alternatives that are freely accessible.
• The most important information sources currently at risk of an outage are the bibliographic database MEDLINE (PubMed) and the study registry ClinicalTrials.gov. This document is intended to provide support in the event that access to these two services is no longer possible.

• The MEDLINE (PubMed) archive would remain accessible in the event of an outage and could be searched via fee-based platforms such as MEDLINE (Ovid) or free services such as Europe PubMed Central (PMC). Any outage would only affect future data, so the resulting gap would only appear slowly, but would particularly affect the most current research. However, there is also uncertainty here as to the extent to which existing cooperation agreements between PubMed/PubMed Central and Europe PMC will continue to be upheld.
• Currently, other freely accessible bibliographic databases or their combination cannot completely replace the content of MEDLINE (PubMed). A combination of a medical bibliographic database such as Embase (fee-based) and the Crossref data pool (e.g., via the "Lens" search interface) would be a partly fee-based and more resource-intensive approach.
• However, the main problem lies less in the basic coverage than in the retrievability of the references. The search functions, extensive metadata and search filters (e.g., "Hedges filters") for MEDLINE (PubMed), which have been developed over decades by information specialists, cannot be transferred to other platforms, or only to an insufficient extent, and therefore cannot be easily replaced.

• The ClinicalTrials.gov archive will continue to be accessible via the WHO ICTRP search portal. However, the portal’s entries are only an aggregate of the ClinicalTrials.gov entries. Information on study results and the documents stored in ClinicalTrials.gov is completely missing in the WHO registry. In a next step, further search steps would have to be taken when accessing the WHO registry (e.g., via the Internet Archive WayBack Machine) in order to identify the associated study documents. Nevertheless, reliable access to or completeness of related study documents from an internet archive would not be guaranteed.
• Future studies can be registered in national (e.g., DRKS) or international study registries (e.g., ISRCTN). The European study registry CTIS is still of great importance for drug studies and should also be opened up to voluntary registration of other study types or research questions, for example.

PubMed is a bibliographic database operated by the National Library of Medicine (NLM). It includes references from the fields of medicine, biochemistry, genetics and molecular biology, pharmacology, toxicology and pharmacy. It offers free access to MEDLINE and PubMed Central (full texts), among others.

MEDLINE was developed in the 1960s and includes citations from more than 5200 scientific journals worldwide. In total, the PubMed database contains more than 38 million citations and abstracts of biomedical literature.

Alternatives to PubMed (MEDLINE)

Several databases, some of which differ significantly in their characteristics, are possible alternatives to MEDLINE (PubMed). Appendix A provides an overview of these possible alternatives. The selection was based on the data pools available, the search functions and availability. These alternatives were primarily identified via https://www.searchsmart.org.

Table 1 shows the extent to which the data pools of the Embase and Crossref databases overlap with MEDLINE, based on current journal lists from these two databases.

It was also examined which databases provide access to the MEDLINE archive with sufficient search functions (e.g. Boolean operators, bulk export). This resulted in the following recommendations, which are presented in Table 2.

As Crossref¹ pools content from leading scientific publishers, the database has a comprehensive data set. However, Crossref's search interface is not suitable for conducting systematic searches, which means that Crossref data is currently searched via so-called academic search engines².

Initial analyses show that of these academic search engines, only Lens meets the requirements for conducting systematic searches. Lens supports both word distance searches and the use of truncation. But there are still limitations: The quality of metadata, such as language information and abstracts, is in need of improvement. In addition, validated study filters are presently not available. The search interface also does not always work reliably and could be optimized.

Paid platforms such as MEDLINE (Ovid) or free services such as Europe PMC would at least allow users to continue searching the MEDLINE (PubMed) archive. However, as new data added to MEDLINE (PubMed) would no longer be included in the future, gaps would only become apparent when references to current publications could no longer be found. However, it is not yet clear to what extent existing cooperation agreements between PubMed/PubMed Central and Europe PMC will remain in place.

Other freely accessible bibliographic databases can currently not replace the information provided by MEDLINE (PubMed) in its entirety. At present, a combination of a medical bibliographic database such as Embase (subject to a fee) and the Crossref database (e.g. via the ‘Lens’ search interface) would be an alternative that is partly subject to a fee and involves greater effort.

Another key aspect in addition to covering the MEDLINE (PubMed) database as completely as possible is the fundamental retrievability of references. The search functions and extensive metadata available in PubMed, as well as the search filters developed over decades by information experts outside the NLM (e.g. the ‘Hedges filters’), cannot be transferred to other databases or platforms, or only to an insufficient extent, and therefore cannot be simply replaced.

ClinicalTrials.gov is the US registry for registering clinical trials and reporting their results. It was established by the FDA Modernization Act of 1997 and went online in 2000.

Initially, the area subject to regulation only covered the registration of studies within the FDA’s remit. However, since its establishment, the registry has also been open to the voluntary registration of other studies. As a result of further regulatory developments, the registry has expanded to become the largest and most important of its kind in the world. Over half a million studies are currently registered there. In addition, it offers extensive search functions compared to other study registries and, in addition to detailed results, also contains other relevant documents such as study protocols and statistical analysis plans (SAPs).

Alternatives to ClinicalTrials.gov

Possible alternatives to ClinicalTrials.gov are shown in Appendix B. Registries that were primarily designed as national registries were not considered. In the event that ClinicalTrials.gov is no longer available, the following recommendations for searching study registries apply:

A review of the WayBack Machine³ showed that archiving takes place with a delay of up to 10 months here. Furthermore, sample tests showed that ClinicalTrials.gov pages that were archived according to the Internet Archive were not displayed. Reliable access thus appears impossible. The WayBack Machine should therefore only be used in combination with other sources. If the API interface of ClinicalTrials.gov (https://clinicaltrials.gov/data-api/api) remains available, study information could be obtained via this source.

If access to ClinicalTrials.gov were restricted, the archive would still be accessible via the WHO ICTRP Search Portal. However, this is only an aggregate of the ClinicalTrials.gov entries. The information on study results and the documents stored in the registry are completely missing at this access. The information contained therein would only be accessible through additional search steps (e.g. via Internet Archive WayBack Machine).

Future registration of new studies could be carried out via national registries, such as the German Clinical Trials Register (Deutsche Register Klinischer Studien [DRKS]), or via international study registries (e.g. ISRCTN). The European clinical trials register CTIS would be an alternative for drug trials. However, CTIS is not yet open for voluntary registration.

Complete or partial restriction of access to NLM databases would have far-reaching consequences for researchers in medicine and the life sciences in Germany and Europe. Considerations regarding alternatives, e.g. to PubMed and ClinicalTrials.gov, should therefore be shared between the institutions potentially affected, and possible preventive measures should be coordinated, particularly in order to avoid redundancies.

In Europe, the key stakeholders in the healthcare sector should ask themselves how they can counter an outage of PubMed and ClinicalTrials.gov. The following measures would be meaningful:

• Existing databases could take over the tasks of PubMed. For example, Europe PMC or the German Central Library (ZB Med) with Livivo (and potentially the Open Life Science Publication Database [OLSPub] in future), could set up the necessary infrastructure.
• The creation of an interface with professional search functions. Since most of the MEDLINE data are freely available via Crossref, an API interface could be used to create a separate interface and thus overcome the limitations of Lens described above.
• For study registries, the task is to expand and open up existing alternatives (e.g., CTIS) or support central platforms (e.g., ISRCTN).