Evidence-based practice (EBP) represents an amalgamation of research evidence, clinical experience and expertise, and patient values or preferences in the process of clinical patient care [1], as shown in Figure 1. This study introduces a contextual approach to research evidence modeled using design science research (DSR) [2] so health care practitioners can improve EBP.

A positive correlation was found between clinician’s behavior and utilization of research evidence from information systems as identified by a study involving 439 nurses and physicians from public and private hospitals [3]. The study, however, concluded that more support is needed for health care practitioners to use research evidence systems.

There is a myriad of evidence-based clinical information resources across various publisher platforms and use of these resources varies from practitioner to practitioner as illustrated in Figure 2.

Electronic resources (e-resources) themselves are classified as journals, books, databases, and clinical decisions support references, and each classification has many associated publishers, vendors, and electronic service providers. Being aware of and retrieving information from all these e-resources was challenging for clinicians who needed information for patient care, medical research, and professional growth [4].

At the turn of the 21st century, this problem was solved to an extent by federated search systems; however, they were unreliable and slow as they connected with each publisher platform in real time, causing users to wait several minutes to see results or miss results entirely when connectors failed [5].

Evidence-based clinical information is traditionally licensed content inaccessible to Web search engines, so the information retrieval gap was satisfied by Web-scale discovery (WSD) services starting with OCLC’s (Online Computer Library Center) WorldCat Local in 2007, Summon from Serial Solutions in mid-2009, EBSCO Discovery Service (EDS) in 2010, followed by Ex Libris Primo Central and Innovative interfaces Encore Synergy [6].

WSD service–based solutions offer user-friendly search interfaces, relevance ranking, and large, centralized indexes, allowing rapid, simultaneous searching [7]. Figure 3 draws upon Figure 2 and illustrates how a WSD solved the issues of fragmented and unknown e-resources by indexing licensed content metadata and making it retrievable by clinicians using a single interface on the screen. By pre-harvesting content from myriad databases into a single index, WSD tools improve on federated searching tools’ speed, de-duplication abilities, relevancy rankings, and the amount of content that can be accessed [8]. WSD services are used worldwide; for example, the EDS is used in over 100 countries across 11,000 institutions [9].

Focusing on the medical sector, a health science research comparing WSD from 3 vendors concluded that none of the services were overwhelmingly more effective in retrieving relevant information, placing them at 50% to 60% likelihood an average user would find adequate resources [7].

When introducing WSD, a study by Hoy uncovered various disadvantages that needed to be addressed further through new research [10], namely, (1) searches produced too many results overwhelming users, (2) object types and formats caused confusion, (3) users experienced noise, and (4) users needed help using limiters to retrieve specific information.

There are also challenges in offering the same set of e-resources to all clinician groups, as a study [11] comparing the information delivery needs of physicians and nurses found significant discrepancies with wide gaps in behavior and motivation as physicians accessed different online resources from nurses for clinical practice.

Articles from the Journal of Medical Internet Research since 2010 discussed research evidence and barriers faced by health care professional to find, appraise, and apply emerging evidence at the point of delivery [12]. As a means to improve EBP adoption rates, health care organizations adopt several strategies such as local consensus processes, distribution of education material, outreach visits, and reminders; however, lack of time, perceived difficulty, and nonintuitive platforms are cited as issues [13]. Web-based knowledge resources had significant impact on access to evidence-based information, which positively correlates with the growth in usage of WSD services over the 2010 decade [14].

It can be deduced by comparing Figure 2 and Figure 3 that although WSDs gave health care practitioners a single point of entry for information retrieval, the array of aggregated content becomes problematic for individuals and clinical groups requiring context they would otherwise get by using the various content platforms directly.

This research aimed to develop and evaluate a model [15] to improve information retrieval from WSD services so health care practitioners can better adopt EBP by satisfying research evidence needs. The model introduces a context layer to existing WSD services to reduce the ambiguity of information by pre-capturing and integrating clinician’s context during the information retrieval process [16].

The contextual layer improves WSDs having a computer science perspective to a knowledge formulation or acquisition system based on the theory of information need for information retrieval by Cole connecting information to knowledge [17]. Figure 4 focuses on the WSD part in Figure 3 and illustrates the improvement by adding a context layer between the user and the WSD that meets requirements listed in Table 1 derived from the literature referenced above.

This section reviews prior work including problems and findings from practice relevant to this research. It also explores prior knowledge and artefacts attempting to solve similar problems to scope knowledge gap. Finally, a literature-based conceptual framework is developed to guide this research.

Using a theory of information need, this research looks to approach developing the solution artefact differently compared with previous attempts discussed in the literature review. Health care practitioners use evidence-based clinical information to deliver patient care, conduct clinical research, and fill knowledge gaps. The WSDs are central to connecting clinicians’ information needs to knowledge, and the literature reveals the 2 primary functions of WSDs, which are aggregation of resources or knowledge and facilitate searching using algorithms. A contextual layer is needed to improve information retrieval, and for this we look to Cole’s theory of information use. Figure 5 shows how the following types are layered inward such as peels of an onion:

The fulfillment of the prefocus, focusing, and postfocus inward journey of the user will help form the contextual framework necessary to improve information retrieval from WSDs. Applying these attributes to the context model is central to incorporating a user-oriented lens to this research as highlighted by Cole [17].

While reviewing implementation of WSD services, it is evident that these systems can be customized; however, the context is limited to the institutional level. A key aim of this research is to improve the current systems and offer personalization or context at the clinical group and user level as illustrated in Figure 6.

The study adopted a constructivism paradigm to maintain an appropriate research philosophy [24]. It is a paradigm that encourages intuitive thinking and guesswork from researchers who should discover and learn principles, facts, and concepts for themselves.

This research will use DSR methodology (DSRM), which falls under the constructivist paradigm to develop a solution artefact model as part of a larger context [25] to solve the problems described earlier in this study. In the case of this research, the artefact or context model created to address the problem [26] is an innovative approach that applies context to information retrieval from WSDs to improve research evidence as part of EBP.

Referring to the knowledge contribution framework in Figure 7, this research falls under the improvement quadrant [27] as it addresses a known problem clinicians and health care professional face while retrieving evidence-based information from WSDs for EBP by developing an innovative artefact model prescribing a contextual approach to retrieving results from WSDs.

The use of DSRM for improvement is a proven principle based on several case studies [2].

The DSRM illustrated in Figure 8 consists of 6-step nominal process sequences that interact iteratively based on findings from subsequent processes [2]. During step 1, the problem is identified along with motivation for a solution, described in the Introduction section of this study.

This is followed by defining solution objectives (Aims section of this study), which serves as inputs to the design and development for creating the artefact. Versions of the artefact are developed through iterative prototyping [28] and used to solve the problem in the demonstration step.

The evaluation step involves determining the artefact’s usefulness based on validity, utility, quality, and efficacy. This research uses the technical risk and efficacy strategy [29] for iterative artificial formative evaluations. Naturalistic summative evaluation via convergent interviewing [30] was used to determine the validity and utility of the artefact with 5 health care practitioners who each were practicing actively for over 15 years. A confirmatory focus group [31] consisting of 5 doctoral level health researchers doing academic research within digital health care for at least 3 years was also conducted, and the transcribed content was analyzed quantitively [32] to ensure significant discussion to support the group’s conclusion.

Findings from this study are communicated using recommendations by Gregor [27] to contribute to the existing body of knowledge.

Up to this point, this research study has described the need for EBP and how WSD services solved the problem of accessing research evidence as illustrated in Figures 2 and 3, but there are problems with the current services. Figure 4 proposed an improvement by applying a contextual layer to Cole’s model of a user-centered system instead of a computer science–focused system as illustrated in Figure 5. Figure 9 draws on gradual development of this theory to propose a contextual model to integrate with existing WSD services.

Starting from the rightmost end of the illustration, a user first borrows a frame for information retrieval for prefocus. For WSD services without the contextual model, the frame is an institutional level context that was configured as part of the WSD implementation. For WSD services that implement the contextual model, the frame can be further personalized to the user by allowing them to borrow a peer’s frame. Focusing deeper, a user’s focus formulates as part of the user’s own information needs until an information need is fully formed as postfocus to capture the context and submit it with the query to retrieve results that are applicable to the user’s context.

Developing the context, however, is a prefocus step where the model allows institutions to iteratively develop clinical group templates to borrow in the group template store by analyzing resource needs if a WSD was not available. These initial templates satisfy the adopt group context requirement of the model to help create a scope for information retrieval.

Individual users then have the option to personalize the template and define a more specific user context, which is stored in the user preference data store. This introduces persistence for context, so users do not have to define their context every time they initiate a new session on the WSD. The context is applied as soon as the system recognizes the user.

Data collection and analysis of the model was done through observation and refinements based on iterative prototyping [28]. Instantiations of the model were developed on the EDS [33] using its application programming interface (API) to iteratively prototype the context layer. A combination of Web technologies such as HTML, CSS, and JavaScript was used for the mobile first front-end. PHP was used for server-side, whereas Google Firebase [34] was used for authentication and storage.

On the basis of observations and comparison of results with and without the context layer, the model and subsequent instantiations were refined several times until it was observed that the model had an impact on information retrieved from the WSD to satisfy R1 to R5.

Once observations showed sufficient evidence that the prototype as a context layer satisfied R1 to R5 for information retrieved from a WSD compared with its noncontextual state, the instantiation was evaluated in the real world with select health care practitioners using convergent interviewing [30] for a summative evaluation followed by a confirmatory focus group consisting of health researchers.

As per DSRM, demonstration was used to solve the problems, observe results, and develop an instantiation or prototype of the context model [2].

The first step is to recognize the user. The developed prototype allows users to sign in using providers such as GitHub, Twitter, or Facebook shown in Figure 10. Once the user authorizes the instantiation to recognize the user, it checks to see if the user has already borrowed a frame or if the default institutional frame should be applied.

The user is presented with all the filters that are available through the WSD API, and the user context is applied automatically if the frame was borrowed before. If the user logs into the instantiation for the first time, none of the filters are preselected, and the default frame is the institutional frame. User can choose to define their context or borrow a context from predefined frames. The prototype shown in Figure 11 includes frames for a general practitioner, nurse, physiotherapist, medical researcher, and psychologist. The model allows institutions to define and present any number of frames to borrow.

The settings enabled through the EBSCO WSD to define the context are databases, source type, subject, publication, and publisher. These settings will be different for WSD from other vendors.

It is not necessary for a user to define a full context right away, and this can be a gradual process as the prototype will store the selections for future reference allowing for gradual refinement of the context.

The screenshot in Figure 12 shows a comparison between information retrieved from the prototype with the context layer (left) versus results without the context layer (right). This example borrows the general practitioner frame to submit the query “fatty liver cure.” The example with the context layer (left) retrieved 5101 results, and the context is visible under the active facets section. The same query without the context layer (right) returned 28,000 results using the institutional frame. The type of information retrieved is presented on screen for qualitative evaluation.

“The technical risk and efficacy evaluation strategy emphasizes artificial formative evaluations iteratively early in the process but progressively moving toward summative artificial evaluations” [29].

A key risk in the evaluation was to determine if a contextual layer could technically be implemented to improve WSD services, so the initial version of the model was developed using the theory of iterative prototyping to allow observations and refinements. Once data collection and observations showed that WSD information retrieval using the context layer satisfied R1 to R5 as part of artefact formative evaluations, a naturalistic summative evaluation was conducted using qualitative and quantitative research methods with health care professionals and health researchers, respectively.

Using a combination of artificial formative evaluations based on iterative prototyping followed by naturalistic summative evaluation using convergent interviewing and confirmatory focus group, the context model satisfied all requirements R1 to R5 of this research.

This research reviewed the state of EBP by health care practitioners and identified that the research evidence dimension was a significant barrier because of the vast myriad of resources. WSD services facilitated research evidence by indexing most licensed content while offering practitioners a single point of entry into research; this solved the problem to a certain degree.

The problems with this approach as identified in the literature was the lack of relevancy to the medical field, creation of noise, unprioritized resources, and missing user context. These problems were synonymous with information retrieval that were developed with a computer science approach focusing on systems and algorithms instead of user need.

On the basis of the theory of information need by Cole, this research designed and evaluated a contextual layer or artefact model to apply to WSD services to improve them. Using DSRM principles, this research used iterative prototyping to observe, design, and refine the model and its prototype against the EBSCO WSD using its APIs.

Once the prototype demonstrated that it satisfied requirements R1 to R5 during a series of artificial formative evaluations, 5 health care practitioners were selected for convergent interviewing for a qualitative naturalistic summative evaluation.

Outcomes from this research provided very positive views about the artefact as all health care practitioners who participated in the field experiments found the proposed approach useful, compatible to their own practices, and added value to WSD services to improve EBP.

A quantitative naturalistic summative evaluation was also performed by forming a confirmatory focus group consisting of 5 health researchers who discussed and concluded that the artefact instantiation met all the requirements. The discussion was recorded and transcribed for content analysis to identify descriptors or codes and associate them to objects or requirements to form Table 4 for a positive hypothesis that there was evidence of significant discussion across the requirements to justify the conclusion.

Evaluation is one of the crucial elements of DSR as it provides feedback to further improve the artefact and bring it to a state where its utility, quality, and efficacy are validated [26].

Prior research [29] shared strategies for DSR evaluation, namely, quick and simple, human risk and efficacy, technical risk and efficacy, and the purely technical artefact strategy. The choice of evaluation depends on the functional purpose of the artefact, which in this research is formative as it aims to improve the outcomes of the process under evaluation. It is also summative as the outcomes will be judged to what extent they match expectations. The other factor for choice of strategy is the evaluation paradigm: if it is artificial or naturalistic.

In this dimension, the paradigm is both artificial and naturalistic as the context artefact is evaluated using iterative prototyping to determine if the context is feasible with a WSD as part of its development. This guided the development in an incremental and iterative fashion to address the research problem. Here, the prototype design and development effort is used to validate or invalidate the theory [28].

The evaluation concluded that the context layer or artefact model had validity and utility as an artefact to solve problems with WSD for EBP. Rigorous research in DSR [29] showed that formative evaluation is necessary to identify weaknesses and areas of improvement during artefact development, and summative evaluation is important to conduct the DSR study by outlining overall utilities or efficiencies and potential benefits of the proposed artefact. Considering artificial evaluation via conducting iterative prototyping and naturalistic summative evaluation, the technical risk and efficacy evaluation strategy was an appropriate fit for this research.

Further research into the contextual approach is recommended to evaluate the artefact using the human risk and effectiveness strategy.