Barriers and facilitators to clinical information seeking: a systematic review

Abstract

Objective

The study sought to identify barriers to and facilitators of point-of-care information seeking and use of knowledge resources.

Materials and Methods

We searched MEDLINE, Embase, PsycINFO, and Cochrane Library from 1991 to February 2017. We included qualitative studies in any language exploring barriers to and facilitators of point-of-care information seeking or use of electronic knowledge resources. Two authors independently extracted data on users, study design, and study quality. We inductively identified specific barriers or facilitators and from these synthesized a model of key determinants of information-seeking behaviors.

Results

Forty-five qualitative studies were included, reporting data derived from interviews (n = 26), focus groups (n = 21), ethnographies (n = 6), logs (n = 4), and usability studies (n = 2). Most studies were performed within the context of general medicine (n = 28) or medical specialties (n = 13). We inductively identified 58 specific barriers and facilitators and then created a model reflecting 5 key determinants of information-seeking behaviors: time includes subthemes of time availability, efficiency of information seeking, and urgency of information need; accessibility includes subthemes of hardware access, hardware speed, hardware portability, information restriction, and cost of resources; personal skills and attitudes includes subthemes of computer literacy, information-seeking skills, and contextual attitudes about information seeking; institutional attitudes, cultures, and policies includes subthemes describing external individual and institutional information-seeking influences; and knowledge resource features includes subthemes describing information-seeking efficiency, information content, information organization, resource familiarity, information credibility, information currency, workflow integration, compatibility of recommendations with local processes, and patient educational support.

Conclusions

Addressing these determinants of information-seeking behaviors may facilitate clinicians' question answering to improve patient care.

INTRODUCTION

Clinicians commonly face clinical questions that arise during patient care, yet search for answers to only about half of these questions and find answers for only a subset of questions pursued.^1–4 In their searching (information seeking), clinicians rely on a variety of resources including paper texts, websites of varying credibility, other humans (ie, curbside consultations), and purpose-built electronic knowledge resources.^5–7 Identifying approaches to facilitate efficient, successful information seeking (ie, finding an answer) at the point of care is essential for evidence-based clinical practice.^8–10 Optimizing the use of electronic knowledge resources such as UpToDate, Micromedex, and Epocrates is particularly important given their growing prevalence in clinical practice. Studies confirm that electronic knowledge resources improve outcomes of knowledge, behaviors, and patient effects; yet effects vary widely from study to study, suggesting that different designs and implementation strategies may have different effects.¹¹

A synthesis of evidence identifying barriers to and facilitators of point-of-care information seeking and knowledge resource use would help clinicians, developers, informaticians, and purchasers anticipate resource features and implementation strategies that would enhance clinicians' efforts. Although reviews of related topics, such as evidence-based medicine and guideline implementation,^12–14 have touched on barriers to point-of-care information seeking and the use of various information resources, we are not aware of reviews that examine these issues in depth. To this end, we aimed to identify barriers to and facilitators of information seeking and knowledge resource use at the point of care as described in published research studies, through a systematic review. We emphasized, and specifically sought evidence regarding, the use of electronic knowledge resources.

MATERIALS AND METHODS

This study was part of a systematic review⁷ of electronic knowledge resources and point-of-care information seeking that was planned, conducted, and reported in adherence to standards of quality for reporting systematic reviews.¹⁵ We first created a comprehensive list of elements reflecting barriers and facilitators as identified in original qualitative research studies, then synthesized these elements into a model representing key determinants of information-seeking behaviors.

Search strategy

The full search strategy has been previously reported.⁷ Briefly, on February 14, 2017, with assistance from an experienced librarian, we searched MEDLINE, Embase, PsycINFO, and the Cochrane Library Database for comparative studies of electronic knowledge resources and information seeking at the point of care, published after January 1, 1991. We reviewed the bibliographies of previous reviews to identify omitted studies.^1,11,16,17 No exclusions were made for language.

Article selection

In a previously reported scoping review of these studies,⁷ we identified 85 as using quantitative or qualitative methods to examine barriers to and facilitators of knowledge resource use or point-of-care information seeking. Two reviewers (CAA and DAC) worked independently to screen the full text of these studies for inclusion in the present review. We included all original studies that used rigorous qualitative methods (operationally defined as explicitly stated qualitative methods for data collection and analysis) to investigate barriers to and facilitators of point-of-care information seeking and the use of electronic knowledge resources. We made no restrictions on specific qualitative methods, but excluded studies that reported “themes” without defining the method by which such themes were derived (ie, no analysis plan).^18,19 We anticipated including quantitative studies (eg, surveys in which respondents rated or ranked prespecified barriers), yet ultimately excluded these because they were context- and resource-specific and used only predetermined barrier or facilitator options, and thus did not support generalizable findings. We conducted an ad hoc review of such studies and found that their conclusions largely mirrored those derived from the included qualitative studies.

We defined clinicians as practitioners or students in a health-related field with direct responsibility for patient-related decisions. Clinicians included but were not limited to physicians, physician assistants, nurse practitioners, certified nurse anesthetists, pharmacists, dentists, midwifes, and psychologists. We defined barriers and facilitators broadly as characteristics of the clinical environment, the clinician (eg, skills or attitudes), or the knowledge resource that prevented or promoted answering clinical questions at the point of care. We defined electronic knowledge resource as a “computer-based resource comprising distilled (synthesized) or curated biomedical information that allows clinicians to select content germane to a specific patient to facilitate medical decision-making.”⁷ All inclusion disagreements were resolved by consensus.

Data extraction

Two reviewers (CAA and DAC) used a standardized data abstraction form to independently extract data from all included studies. We extracted information about key study characteristics (participants, methods, resources studied, and potential conflicts of interest). We appraised qualitative study methods using the framework described by Popay et al¹⁹ as adapted by Hatala et al,²⁰ namely domains of theoretical adequacy, reflexivity, responsiveness, sampling, thick description, data quality, auditability, analysis, and relevance. Reviewers resolved by consensus all disagreements on key characteristics and quality appraisal. Interrater reliability of study quality appraisals was generally high across the separate domains (Supplementary Appendix Table 1).

These reviewers inductively created and iteratively revised a taxonomy of elements describing barriers and facilitators, adding and merging elements throughout the abstraction process (Supplementary Appendix Table 2). As the taxonomy evolved, previously reviewed studies were recoded. A barrier or facilitator was counted as present in a given study if either reviewer identified this element.

Data synthesis

Three reviewers (CAA, DAC, LAM) worked iteratively and collaboratively to group specific barriers and facilitators into related clusters. We sought common and inter-connected concepts among these clusters and synthesized these into themes representing key determinants of clinicians’ point-of-care information seeking. We gave greater weight to studies that employed stronger methods; for example, we used several rigorous and comprehensive studies as the starting point in developing our taxonomy of themes. Acknowledging the influence of researchers’ experiences on this synthesis, we note that CAA is a general internist with formal clinical informatics training, DAC is an education researcher with expertise in computer-assisted learning, and LAM is an information scientist with experience teaching and researching evidence-based medicine.

RESULTS

Trial flow

We identified 10 811 potentially relevant studies: 10 799 from our literature search and 12 by examining related reviews. From these we found 45 studies meeting inclusion criteria (Figure 1).^21–66

Figure 1.

Study flow.

Study characteristics and quality

Data collection involved 1-on-1 interviews (26 studies), focus groups (21 studies), ethnography (6 studies), analysis of logs or written data (4 studies), and usability studies (2 studies). Table 1 summarizes the characteristics of included studies. For analysis, 15 studies utilized grounded theory, 4 used a case study approach, 1 used phenomenology, 16 studies used “thematic analysis” without citing a specific method, and 8 studies used a vague analysis method (Table 2). Most studies were conducted in the context of general medicine (n = 28) or internal medicine specialties (n = 13). Practicing physicians (n = 26), postgraduate physician trainees (n = 11), and medical students (n = 11) were most frequently studied. The quality of studies was highly variable: 19 (42%) studies were appraised as meeting high quality for 3 or fewer (of 9) domains, whereas 8 (18%) met high-quality criteria on 7 or more domains (see Supplementary Appendix Table 1). Three (7%) studies reported potential conflicts of interest. Funding sources were reported in 37 (82%) studies.

Table 1.

Characteristics of included studies (N = 45)

Data collection method a
Focus group	21 (47)
Interview	26 (58)
Ethnography	6 (13)
Log/written	4 (9)
Usability	2 (4)
Participants
Physician	26
Resident/fellow	11
Medical student	11
Nurse practitioner or student	7
Physician assistant	1
Pharmacist	3
Other	1
Context/topic b
General medicine	28
Medical specialty	13
Emergency medicine	2
Obstetrics/Gynecology	1
Pediatrics	2
Other/NOS	8
Analysis framework
Grounded theory	15
Thematic	16
Case study	4
Phenomenology	1
Indeterminate/NOS	8
Quality appraisal (high quality)
Theoretical adequacy	25 (56)
Reflexivity	3 (7)
Responsiveness	11 (24)
Sampling	17 (38)
Thick description	25 (56)
Data analysis	24 (53)
Auditability	27 (60)
Analysis methods	25 (56)
Relevance	24 (53)

Values are n (%) or n.

NOS: not otherwise specified.

^aSome studies utilized multiple data collection methods.

^bSome studies concerned multiple clinical contexts.

Table 2.

Key characteristics of included studies

Author (Year)	Participants Type, no. enrolled	Data Collection	Analysis	Topic/Context	Time	Access	Pers-SA	Instit	Features
Wood (1995)21	MD; N = 62	Interview	Grounded, Other/NOS	General Medicine	X	X	X		Cur, Org, Vol
Forsetlund (2002)22	MD; N = 52	Interview, Focus Group, Ethnography	Thematic	General Medicine	X	X	X	X	Org, Sco
Adler (2003)33	PG	Focus Group	Thematic	Emergency Med	X	X			Opt, Sco
Gosling (2003)58	MD; N = 17	Interview	Grounded	Other/NOS	X	X	X	X	Sco
Prosser (2003)44	MD; N = 30	Interview	Grounded	General Medicine			X	X	Cred
Bryant (2004)55	MD; N = 58	Interview, Focus Group	Case	General Medicine	X
Johnston (2004)62	MS	Focus Group	Grounded	General Medicine	X	X	X	X	Sco
Lapinsky (2004)63	MD; N = 13	Focus Group	Vague	Medical Specialty		X			Cur, Opt, Org, Sco
Chew-Graham (2005)64	MD; N = 24	Interview	Grounded	General Medicine	X		X		Org
Ely (2005)61	MD; N = 48	Interview, Ethnography	Thematic	General Medicine, Pediatrics		X	X		Cred, Cur, Eff, Opt, Org, Sco, Vol
Green (2005)65	PG; N = 34	Focus Group	Grounded	General Medicine	X	X	X	X	Org
McCaughan (2005)66	NP; N = 33	Interview, Ethnography	Grounded	General Medicine	X	X	X		Cred, Vol
Garrett (2006)23	MS, NP-S	Focus Group, Log	Thematic	General Medicine	X	X	X		Org
Tan (2006)24	MD, PG, Pharm; N = 32	Interview	Thematic	Medical Specialty	X	X	X
Zack (2006)25	MD; N = 47	Interview	Grounded	General Medicine		X			Cred, Cur, Opt, Org, Sco
Adams (2007)26	NP; N = 12	Interview	Thematic	General Medicine	X	X	X		Cred, Eff, Org, Pat, Sco, Vol
Pluye (2007)27	PG; N = 26	Interview, Log	Case	General Medicine	X				Sco
Mysore (2009)28	PG; N = 17	Interview	Case	General Medicine				X	Cred, Loc, Opt, Org, Sco
Ajjawi (2010)29	MD, PG; N = 43	Interview, Focus Group	Vague	General Medicine	X		X		Cur, Sco, Vol
Kastner (2010)30	MD; N = 16	Focus Group	Grounded	General Medicine, Medical Specialty	X				Loc, Opt, Org, Sco
Robertson (2011)31	MD, PG; N = 27	Interview	Thematic	General Medicine		X	X		Cur, Eff, Fam, Int, Opt, Org, Pat
David (2012)32	MD, NP; N = 24	Interview	Phenomenology	Medical Specialty	X	X			Eff, Fam, Org, Sco
Davies (2012)34	MS; N = 140	Focus Group, Written	Grounded	Other/NOS		X	X	X	Eff
Hains (2012)35	MD, PG, NP, Pharm; N = 61	Interview, Focus Group, Ethnography	Thematic	Medical Specialty		X	X	X	Loc, Opt
Lindgren (2012)36	MD; N = 9	Interview, focus group, ethnography	Vague	Medical specialty	X				Eff, Org
Menon (2012)37	Other; N = 14	Usability	Thematic	Medical specialty		X			Cur, Eff, Opt, Org, Sco
Baudains (2013)38	MS; N = 12	Focus group	Thematic	General medicine		X		X	Cred, Eff, Fam, Opt, Org
Baysari (2013)39	MD, Pharm; N = 20	Interview, focus group	Vague	Pharm		X			Org, Vol
Cook (2013)41	MD; N = 50	Focus group	Grounded	General medicine, medical specialty					Cred, Cur, Eff, Fam, Int, Loc, Opt, Org, Pat, Sco, Vol
Cook (2013)40	MD; N = 50	Focus group	Grounded	General medicine, medical specialty	X	X			Cred, Cur, Eff, Fam, Opt, Org, Pat, Sco, Vol
Hardyman (2013)42	MS; N = 260	Written	Vague	Other/NOS					Eff, Opt, Sco
Henderson (2013)59	MD, NP; N = 6	Focus group	Thematic	General medicine				X
Khalifian (2013)43	MS; N = 6	Written	Vague	Other/NOS		X			Cred, Eff, Opt, Org, Sco
Brennan (2014)45	MD, PG, MS; N = 46	Interview, written	Thematic	General medicine		X			Cred, Cur, Fam, Org
Devine (2014)46	PG; N = 10	Usability	Case	Medical specialty					Cred, Org, Sco, Vol
Eng (2014)47	N = 15	Focus group	Thematic	Medical specialty		X			Cur, Fam, Org
Maggio (2014)48	MD; N = 22	Interview	Thematic	General medicine					Cur, Opt, Sco
Nuss (2014)49	MS; N = 37	Interview, ethnography	Thematic	General medicine		X
Bradley (2015)50	MD, NP, PA; N = 50	Interview	Grounded	General medicine		X		X	Fam
Townsend (2015)51	MD; N = 14	Focus group	Grounded	Medical specialty	X		X
Maggio (2016)52	MD; N = 38	Interview	Other/NOS	Other/NOS	X	X			Cur, Pat
Schuers (2016)53	MD, PG; N = 35	Focus group	Vague	General medicine	X		X		Cred, Cur, Eff, Fam, Loc, Opt, Org, Pat, Sco
Templeman (2016)54	MS; N = 30	Interview	Grounded	Other/NOS		X			Cred, Eff, Fam, Int, Org, Sco
Witt (2016)56	MS; N = 34	Focus group	Vague	General medicine		X	X	X	Eff, Pat
Twiss-Brooks (2017)57	MS; N = 68	Interview	Thematic	Anesthesiology, emergency medicine, laboratory medicine and pathology, general medicine, medical specialty, obstetrics/gynecology, pediatrics, psychiatry, surgery	X	X			Fam

Access: accessibility; Cred: resource credibility; Cur: resource currency; Eff: Resource efficiency; Fam: resource familiarity; Features: features of specific knowledge resources; Instit: institutional culture, attitudes, and policies; Int: resource integrated with workflow; Loc: compatibility with local practices; MD: physician; MS: medical student; NP: nurse Practitioner; NP-S: nurse practitioner student; Opt: resource optimization for question answering; Org: resource information organization; PA: physician assistant; Pat: patient education support; Pers-SA: personal skills and attitudes; PG: resident/fellow; Pharm: pharmacist; Sco: resource scope; Vol: resource information volume.

Key determinants: specific barriers and facilitators of information seeking

Our model of key determinants (Table 3) encompassed 5 domains, each with several subthemes. In the text below we present a synthesis of key findings. Representative quotes supporting and elaborating upon these key findings can be found below and in Table 3. The number of quotations does not necessarily reflect the strength of the subtheme in the literature; quotes were selected to illustrate the breadth of concepts represented in a given subtheme.

Table 3.

Model with selected representative statements and quotations for barriers and facilitators of point-of-care information seeking

Key Determinant Domain and Subtheme	Representative Statements and Quotes
Domain: time (n = 22)
Time availability	• B: As a result of the nature of hospital work, time constraints are also a significant barrier to the greater use of [knowledge resource] at the point-of-care.24
	• B: Looking for information was epitomized as time consuming and frustrating.55
	• B: Time constraints kept them from looking for information.21
Urgency of information seeking need	• F: “Some cares are so important that you just have to [search for information].”22
Efficiency of information seeking	• B: “I spent over an hour looking for an answer and came up with nothing useful.”61
Tracking deferred questions	• B: If unable to respond to a question as a clinician scenario unfolded, residents often deferred the question to a later time. However, they rarely pursued these postponed questions. … [they] lamented the lack of an adequate system to track them.65
Domain: resource accessibility (n = 30)
General	• F: “What’s most important is that it is easily accessible.”22
Hardware	• B: Even if the computer was free for the moment, they feared that the computer would be needed for a clinical function.33
	• B: “We don’t have enough computers and the ones we do are really old.”35
Computer/Connection speed	• B: “There are some days where you … think you’ll just check something, [but] you’ll actually have to give up because it’s just too slow.”64
	• B: Residents also lamented inferior technology, including outdated hardware, slow Internet connections, firewall restrictions, and inability to make printouts.65
	• B: “You haven’t got time to wait for things to download for 20 minutes.”66
Information restriction	• B: “If it asks for a password I’m stuffed because I wouldn’t have one.”66
	• B: “I’m pulmonary clinical care, and one of the major [journals is] CHEST. And my hospital library made a decision that they will only have access to the archives, so I can only get access to papers that are a year old or older.”52
Cost (include institutional payment)	• B: “It would depend on how much I have to pay. If it’s affordable for a student, then I’ll be willing to pay.”54
	• F: “Most of the time I go to [commercial resource] on my computer. The facility pays for that, so I utilize it.”50
Portability	• B: Some expressed a strong desire to use portable electronic devices while “wandering around the hospital” as the ease of transportation would allow them to use online sources “more frequently”.24
	• F: The portability of the tablet … was viewed as helpful in enabling users to carry their tablets around and access information anywhere.56
Domain: personal attitudes and information-seeking skills (n = 19)
General information seeking skills	• B: “The majority don’t feel at all comfortable about their searching skills. … I don’t see ordinary [General Practitioners], in their day-to-day clinical practice, trying to practice evidence-based medicine, doing what I would call a “proper” search.”55
	• B: Residents professed difficulty articulating “answerable questions” and translating them into effective search terms and strategies.65
Computer literacy	• B: A lack of [computer] skills meant that computerized protocols were inaccessible for a large number of nurses.66
Don’t know where to look	• B: “It's hard to know which [resource] is the most useful or which one's recommended, so you wind up wasting your time trying everything.”29
Enlist third party for help	• B: “Sometimes it’s easier to refer the patient than to take the time to look up the information, even if it’s something that I know I could manage, but I just don’t have the time to deal with it.”41
	• B: “I see a role for the librarian to do all the work for us, basically. It’s ‘Why have a dog and bark yourself?’”55
Attitude about computers/technophobia	• B: “A lot of people use time as being an issue and in actual fact they’re technophobic themselves and actually cannot get through that barrier of going to [the computer].”35
Attitude about information seeking with patient present	• B: “There is nothing worse than watching someone word process, so as a sort of social thing I don’t think [seeking information in front of the patient] is a great thing.”64
	• F: “That’s why [commercial search engine] is, to me, so awesome because if I want to show a patient a picture of something, I can get to it so much faster by just [searching].”40
	• B: Trainees worried that consulting information sources … in front of patients might convey inexperience and a lack of competence.31
Domain: institutional attitudes, cultures, and policies (n = 12)
Culture and policies	• B: At the two community hospitals, residents found the culture more inhospitable to meeting their information needs. They encountered a prevailing perception that computers in clinical areas are intended for managing patient data and not for looking up medical information.65
	• B: A lack of organizational recognition [of a knowledge resource] was also identified as a barrier to the use of the platform.32
	• F: “The majority [of GPs] ‘don’t feel at all comfortable' about their searching skills. … If you get an information scientist to do your search, you don't get the unfiltered lot … that's why doing them together would be ideal.”55
	• F: “It would be very useful to me if [a librarian] could help me access information.”22
Local champion	• F: “If you’ve got a guru who in the medical field, the pharmacy field and the nursing field who uses it…the hierarchy uses it in that area, then the local drivers will use it, they fall in line.”35
Domain: knowledge resource features (n = 40)
Efficiency (n = 15)	• F: [Mobile app] was much, much quicker than flicking through [paper resource].42
	• F: “When you have a choice of multiple resources to go to, what you want to know is … very quickly, 1) is the answer here? And you want to spend as little energy possible finding out is the answer here or not; and … 2) how quickly can I get to it through reliable search?”40
	• F: Desirable online information is “not too busy” and succinct.24
Volume of information (length/detail of information per topic) (n = 9)	• F: “I’m trying to address a patient question immediately. It’s good to have [the information] referenced, or the information there, but I need a short synopsis. What do I need to do to the patient right now?”61
	• B: “Sometimes the information that’s present is just too basic. It’s not deep enough and so you’re needing to go to a second or a third source …. And you can use [the resource] effectively often enough that you go back to it, but it likely won’t satisfy your needs every time.”40
	• F: “What you need is really a very clear guide with a main salient point, the main complications you are likely to face and the main thing – what to do about them.”25
Scope (breadth) of topics (topical coverage) (n = 22)	• F: “The reason I used [commercial resource] is because it’s reliable. I know that typically it will be well populated, the information will be there, and if the information is not there then it’s unlikely to be in [local resource] or other places.”40
	• B: Require more specialty-specific content.63
	• B: “[Commercial resource] only presented the information for a patient who isn't stable … My patient was stable.”28
Information organization (n = 25)	• F: “[It is important] that things are well organized and we do not get lost in finding information.”32
	• F: “On websites, very often it is the basics of layout and presentation. If it’s a well laid out website, you are going to look at it and think, ‘gosh, this looks very useful.’ And if it’s easy to move through that’s great.”66
Optimization for answering different types of questions (n = 16)	• F: Their need for different types of content varied depending on the clinical situation. Some wished for brief content to help before they went to see a patient, others wanted direct links to diagnosis or management by disease process, while others liked the current [resource] design based on a mix of symptoms and diagnoses. … The program should be able to offer different kinds of information (diagnostic vs. educational) based on the user’s specific need at that moment.33
	• F: Physicians often preferred different knowledge resources depending on the specific question. For example, while [commercial general resource] was by far the most cited resource overall, most physicians suggested another resource such as [commercial medication-focused resource] when searching for information on drug dosing or side effects.40
	• F: Participants often selected PubMed when they needed more current material … [or if they] wanted to confirm something said by others … [such as when] colleagues quote clinical trial findings. … [Commercial resource] was the resource primarily used when participants needed to answer logistical questions such as ‘What is the half-life of [medication xxx]?’48
Familiarity (n = 2)	• F: “Familiarity makes navigation quicker.”26
	• F: “[Commercial resource] was the best source when I was a resident training and at two in the morning when you wanted a quick answer to get things going, that was the resource. So I got trained into [commercial resource], I guess.”40
Credibility (n = 14)	• F: “I think if it came from a source I trusted, it would stand to reason that they would be…making references clear…and there again…if it is a site that you…respect then it stands to reason that it is updated.”25
	• F: “By being developed by [local institution] we assume that there’s a level of credibility that goes along with whatever answers are in there, and I think that’s probably a fair assumption.”40
	• B: “I didn't find this information anywhere else. I talked about that with a specialist … and she said that in fact it's probably an error in [commercial resource].”28
Currency (n = 14)	• F: Textbooks went out of date very quickly. [Electronic resources did not].21
	• F: Clinicians used hard copy text or reference books ‘less and less frequently’ as information is often out of date by the time it is printed.24
Integration with workflow (n = 3)	• B: “I think that non-integrated things will only ever get used if they are so vastly superior to what’s out there that they form an integral part and you can’t live without them… apart from that, unless something is seamlessly integrated and useful, it will not get used.”31
	• F: Bookmark links or embedded links to [electronic clinical support systems] from patient records were popular among users.26
	• F: “If there were just a button that said ‘Candida blah … [local resource]’ and you could get one click away from a paragraph, ‘What is this bug?’ I’d click it every time, and that’d be the first thing I’d go to.”40
Compatibility with local processes (n = 5)	• B: “Methotrexate wasn't offered in the hospital so it wasn't really an option for the patient.”28
	• F: Sometimes trying to figure out what I need to order here (versus what the current [commercial resource] is telling me to order or do) can be a little different.40
Patient education support (n = 7)	• F: [General Practitioner] and [trainees] also accessed information during consultations, so that they could show patients helpful resources or printout information leaflets to take away.45
	• F: It should have in there whatever patient information kind of stuff we want; both that we can print it off for the patient and give it to them as well as just speak with them.50

All statements are quoted from each original research report. Those enclosed in quotation marks reflect direct quotes from study participants, while those without quotation marks reflect statements by the study authors. n indicates the number of studies reporting this theme (see Table 2 for details on how themes were reported for each study).

B = barrier; F = facilitator; GP: general practitioner.

Time

Time availability and the need to prioritize which questions to pursue were noted as key factors influencing information seeking in nearly all studies, including both old and recent publications. “Time is often too short to go thoroughly into the literature.”²² In addition to the general theme that inadequate time was a key barrier, specific subthemes that influenced the decision to invest time searching for an answer included the priority or urgency of the question, the question complexity, current clinical pressures (ie, time availability), the effectiveness and efficiency or brevity of available resources, and resource accessibility. “There’s a million things you could look up on patients. You have to stratify things.”⁶⁵ Higher urgency encouraged seeking answers, while time was typically viewed as insufficient to tackle questions of greater complexity. “Patients are so much more complex than they were 20 years ago. … There’s actually less time and more pressure.”⁴¹

Accessibility

Closely related to the issue of time was the accessibility of knowledge resources, with subthemes of computer availability, hardware reliability, computer speed, network connection speed, technical restrictions, and resource cost. “What’s most important is that it is easily accessible.”²² The physical location of computing systems (ie, not readily available at the point of care) was identified as a barrier to use in earlier studies, but this barrier was mentioned less frequently in recent years. Hardware reliability issues included hardware failure, syncing limitations and limited storage space for mobile devices, and limited battery life. Technical problems with accessing knowledge resources eroded confidence and limited use. Other technical barriers, such as firewalls, paywalls, and resources not being available in mobile format, also impeded use. The cost of resources also influenced access; decisions regarding resource purchasing or subscription greatly impacted the availability of specific resources. “The facility pays for that, so I utilize it.”⁵⁰

Personal skills and attitudes

Information-seeking skills and attitudes were identified as subthemes affecting search initiation and success. Good general information-retrieval skills increased confidence in search success and efficiency; deficiencies in these skills impeded searching. Sometimes clinicians did not know where to begin their search: “Many clinicians lacked knowledge of appropriate sites to access.”²⁴ In some clinical contexts, enlisting a third party to help answer a clinical question, often a librarian or a consultant, was deemed easier than conducting a search—especially when the clinician self-assessed as possessing inadequate searching skills.

Personal attitudes, such as beliefs that an answer does or does not exist, can or cannot be easily found, or will or will not change clinical management were noted as influencing the initiation of an information search. Additionally, unfamiliarity with or anxiety about computer technologies in general was a frequent barrier to knowledge resource use in early studies; such concerns seemed to fade in later studies and have been replaced by attitudes reflecting preferences for familiar resources. Studies reported conflicting attitudes about using electronic knowledge resources in front of patients: some clinicians believed that this practice erodes patient confidence, while others believed that it helps.

Institutional attitudes, culture, and policies

Institutional attitudes, culture and policies were also reported to influence information seeking. General institutional attitudes and policies toward evidence-based medicine (eg, a culture of evidence-based practice) or toward specific knowledge resources influenced resource use. In some institutions, specific electronic resources were viewed as the “gold standard” reference, often de facto (eg, a site license for a commercial resource), sometimes informally (by tradition), and rarely through stated policy (“We determined as a clinic organization to use [commercial resource] for our primary source of information so that we would all be on the same page.”)⁵⁰ Support from librarians and local champions positively influenced attitudes toward information seeking (“pushing a bit of an information culture.”⁵⁸ Librarian availability for search support helped overcome barriers arising from insufficient time or skills. Clinicians were less apt to seek information in situations that adversely impacted another clinician's workflow, such as limited point-of-care computer availability.

Knowledge resource features

In addition to the key determinants and subthemes noted previously, several studies noted specific features of knowledge resources that promoted or impeded their use. We list these briefly in the following list; see additional supportive quotes in Table 3.

• Efficiency: Many studies reported that clinicians wanted a high likelihood of quickly finding the answer to their questions using a specific resource. “I feel like I’m drowning in a sea of information. I’m afraid of not finding what I’m looking for.”⁵³ Many of the other features listed subsequently ultimately facilitate such efficiency.

• Volume of information: Nearly all studies reported subthemes describing general characteristics of information contained in knowledge resources, such as the volume (overall quantity of information for a given topic) and scope (breadth of topics covered). There was tension between narrow and broad information needs in different situations. In some cases, physicians preferred knowledge resources that provided brief, focused answers rather than an exhaustive well-referenced general resource. “I need a two-sentence answer.”⁶¹ In many other cases, information was perceived as insufficient to answer the question.

• Scope of topics: In general, resources with greater scope (more comprehensive topical coverage) were preferred. Topical coverage (or lack thereof) was frequently mentioned as a reason to use (or not use) a given resource. “Once she discovered that the content was limited, she did not return to [knowledge resource].”³³ However, numerous topics often made finding a specific answer difficult.

• Information organization: Several subthemes addressed issues related to the organization of information within a resource, including navigation, search or indexing functionalities, and clear presentation format. Poor search and navigation were cited as reasons to avoid specific resources. “I think that the search function and the way you find the information is just cumbersome … That’s probably my barrier to using it.”⁴⁰ Conversely, an organized, easy-to-navigate user interface with robust search functionality was identified as an important facilitator of electronic knowledge resources use. Some resources allowed clinicians to enter patient-specific data to tailor the information retrieved.

• Optimization for answering different types of questions: Clinicians desired resources optimized to their information needs. “Residents suggested that their need for different types of content varied depending on the clinical situation.”³³ Different needs (eg, questions related to diagnosis, test interpretation, treatment best practices, or drug prescribing) and contexts (inpatient or outpatient; clinical or teaching) each may require different design optimizations. Algorithms and flow diagrams were helpful for describing care processes, but recommendations that were too prescriptive were disliked. Clinicians preferred clear, concise, and actionable recommendations.

• Familiarity: A clinician’s familiarity with the content, organization, and user interface of a specific resource influenced resource preferences. Familiarity with a given resource often had its roots in medical school or postgraduate training, although it could be augmented during later clinical experiences or classroom-based training.

• Credibility: Clinicians expressed a need to know whether information can be trusted. The reputation of the resource's creator often served as the primary justification for the credibility of the information content. Citations to primary literature (“an evidence-based rationale for recommendations”)⁶¹ also helped to support credibility.

• Currency: Clinicians expected the information content to be current. In general, the currency of information was viewed as a strength of electronic knowledge resources compared with paper resources.

• Integration with workflow: Clinicians favored knowledge resources integrated into their clinical workflow or the electronic health record. Although electronic knowledge resources had highly variable integration in early years, in recent studies clinicians seemed to view them as generally better integrated than nonelectronic resources.

• Compatibility with local processes: Clinicians reported that resources providing information or recommendations incompatible with local care processes were less likely to be utilized, whereas information from local experts or otherwise known to align with local practices was viewed favorably.

• Patient education support: Knowledge resources are often used as part of patient education, and resources that offer information or decision-support tools directed at patients were useful in those contexts.

Interrelationships among determinants of information-seeking behaviors

We noted numerous interconnections among these 5 determinants. For example, the perception of adequate time is influenced by many factors, including personal skills and attitudes, institutional culture and policies, resource availability, and resource efficiency and familiarity. Moreover, the actual time needed to answer clinical questions is affected by features of specific knowledge resources, such as information organization, familiarity, and optimization for specific queries. Attitudes and perceived skills will be influenced by institutional culture and the work environment, including training, librarian support, subscription to familiar and usable resources, and local resource preferences.

DISCUSSION

We identified 45 qualitative studies that explored barriers to and facilitators of information seeking and electronic knowledge resource use at the point of care. We identified 5 key determinants in our model of point-of-care information seeking including perceived time for information seeking, resource accessibility, personal attitudes and skills, the clinical environment, and familiarity with and features of specific knowledge resources. Features influencing the effectiveness of knowledge resources include efficiency, information content, and information presentation. Resources must strike a balance between breadth and depth of information to meet information needs in a variety of clinical situations.

Limitations and strengths

Although several studies were of low methodological quality, we chose not to exclude any studies altogether to maximize the diversity of descriptive themes captured. In general, low-quality studies identified fewer themes rather than novel or dissenting themes. We did give greater weight to studies of higher methodological quality during thematic synthesis. The studies represented a broad range of clinical contexts, participants, and methodologies, which made data coding challenging but enabled a rich synthesis. As we are not attempting a comprehensive synthesis of quantitative outcomes, we view the age of our search as a minor limitation. It seems unlikely that studies published since our search was completed (and thus omitted) would substantially alter the model of key determinants.

Integration with prior research

Our study expands upon a previous review of physician information-seeking behaviors⁶⁷ in its comprehensiveness and by explicitly considering how clinicians actually use knowledge resources. Our findings also extend the findings of previous reviews that focused on interventions to promote adoption of information and communication technologies, by considering the concept of information seeking broadly and by considering in greater detail the key features of knowledge resources.⁶⁸ Our final model of determinants, derived from a comprehensive and detailed analysis of 45 publications, builds on previous work^40,61 by adding a more complete conceptualization of time-based barriers, the impact of personal skills and attitudes, and the impact of institutional culture and policies; and by elaborating on specific features of knowledge resources that facilitate or impede information seeking.

Implications

Our study sheds new light on key issues that enable and impede information seeking. Perhaps most importantly, time constraints represent an ubiquitous barrier to information seeking. We expect clinicians to answer questions, but do not allot the necessary time during clinical encounters. Increased time pressures are compounded by the increasing clinical complexity of patients. While system and policy changes will undoubtedly be required to address this challenge,^69,70 improvements in the accessibility and efficiency of electronic resources will substantially facilitate information seeking at the point of care.

Second, we established that information volume (the amount of information on a subject) is conceptually distinct from information scope (the number of subjects covered by a resource). With respect to volume, there was a tension between too much and too little information within a resource—a subjective assessment based on the clinical context and question at hand. The possibility of dynamically “right-sizing” the amount of information presented to match varying information needs represents an interesting challenge for future research. Information scope presented a similar tension—too many topics negatively affects efficiency in finding answers, while too few decreases the likelihood that an answer exists within the resource. Future knowledge resource designs will need to address the tension between the simultaneous need for breadth and depth to meet clinicians’ information needs—in a variety of clinical contexts. Continued improvements in the search functionality of these resources may help improve the efficiency of more comprehensive resources.

Finally, we have identified several specific features of knowledge resources that clinicians believe will promote information seeking at the point of care, including efficiency, scope, volume, organization, familiarity, credibility, integration with workflow, and compatibility with local processes (see Table 3). In most instances, the features identified in these qualitative studies reflect hypotheses or conjectures about what might improve information seeking, rather than recommendations supported by high-level evidence. Each of these features could be further examined through, for example, human factors and usability studies seeking to advance the design of information retrieval tools.⁷¹ While such evidence accumulates, we suggest these features—synthesized from several empiric studies—constitute a list of best practices that can be implemented immediately (and then tested).

CONCLUSION

Clinicians' information seeking is influenced by time, resource accessibility, personal attitudes and skills, institutional characteristics, and specific resource features. Attention to these factors when developing and implementing knowledge resources may facilitate answering clinical questions at the point of care.

FUNDING

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Uniformed Services University of the Health Sciences, the Department of Defense, or the U.S. Government.

AUTHOR CONTRIBUTIONS

Study concept and design was completed by CAA, LAM, GDF, and DAC. Title and abstract review were completed by CAA and DAC. Full text review and data extraction was completed by CAA and DAC. Data synthesis was completed by CAA, LAM, and DAC. Drafting of the manuscript was completed by CAA and DAC. Critical revision of the manuscript was completed by CAA, LAM, GDF, and DAC.

SUPPLEMENTARY MATERIAL

Supplementary material is available at Journal of the American Medical Informatics Association online.

CONFLICT OF INTEREST STATEMENT

In 2016, LAM received travel funds to deliver a lecture on evidence-based medicine for employees of EBSCO, the parent company of DynaMed; EBSCO did not have any involvement in the conduct of this study. The remaining authors are not aware of any other competing interests.