Physicians Answer More Clinical Questions and Change Clinical Decisions More Often With Synthesized Evidence: A Randomized Trial in Primary Care

Article Figures & Data

Table 1.

Design and Process Principles for DynaMed

Principle	Description
Systematic literature surveillance	Consistent ongoing monitoring of defined information sources considered to have a high yield for primary care
Relevance selection filter	Selection of content limited to clinically relevant information
Validity selection filter	Selection of new content limited to information that meets or exceeds validity of existing content
Concise synopsis	Brief summarization of specific methods and specific outcomes for included content
Direct reference citation	Provision of reference support for specific statements and assertions
Synthesis of evidence	Revision of outlines and overview statements to reflect all evidence summarized
Standardized template	Consistent headings and subheadings for organizing clinical topic outlines
Additional content listings	Listing of external reviews and guidelines

Table 2.

Participant Characteristics (N = 52)

Characteristic	No. (%)
EBM = evidence-based medicine.
Note: Because of rounding, percentages may not total 100.
* Range, 1 to 30 years; mean, 9.7 years; median, 5 years.
† Rated on a scale from 1 (novice) to 5 (expert).
‡ Rated on a scale from 1 (not valuable) to 5 (extremely valuable).
Country
United States	49
Israel	1
Lebanon	1
Pakistan	1
Type of clinician
Physician	49
Nurse-practitioner	3
Medical specialty
Family medicine	45 (87)
Internal medicine	5 (10)
Pediatrics	1 (2)
Women’s health	1 (2)
Number of years in practice*
1–4	23 (44)
5–7	8 (15)
10–15	7 (13)
17–22	8 (15)
24–30	6 (12)
Self-rated familiarity with computer searching^†
2	6 (12)
3	7 (13)
4	25 (48)
5	14 (27)
Self-rated perceived value of EBM^‡
3	1 (2)
4	20 (38)
5	31 (60)

Table 3.

Number of Participants Citing Sources as First and Other Information Sources (N = 52)

Information Source	First Choice When Not Using DynaMed	Listed in “Other Resources Typically Used for Searching”*
PDA = personal digital assistant.
* Not counting listings as first choice. Participants could list multiple resources. Additional resources listed were National Guideline Clearinghouse (4 participants), American Academy of Pediatrics (3 paricipants), American College of Physicians (3), Centers for Disease Control and Prevention (3 participants), Ovid (3), Bandolier (2), National Institutes of Health sites other than National Library of Medicine/PubMed (2), and Clinical Evidence, “derm sites for rashes,” Dogpile, DrPEN, eMedicine, FamilyDoctor.org, “filed articles,” FPNotebook, “materials from CME meetings,” The Medical Letter, PDxMD, Scharr Netting the Evidence, Stat-Ref, and TRIP Database (1 each).
† Web portals were typically library search pages.
‡ Specific journals and textbooks were typically unnamed.
§ PDA was named without specific sources in 3 cases and with Epocrates given as the source in 2 cases.
Google	10	9
American Academy of Family Physicians	8	9
UpToDate	7	8
MDConsult	6	10
Web portal^†	6	4
MerckMedicus	5	0
InfoPOEMs	4	4
Medscape	3	2
MEDLINE (or PubMed)	1	16
Yahoo	1	1
WebMD	1	1
Cochrane Library	0	7
Journals^‡	0	6
Textbooks^‡	0	6
PDA^§	0	5

Table 4.

Results by Per-Participant Analysis

Result*	With DynaMed No. (%)	Without DynaMed No. (%)	No Difference No. (%)	P Value^†
* The n values are numbers of participants.
† The Wilcoxon signed rank test ranks the absolute values of the differences in each paired group, thus comparing the aggregate ranks for With DynaMed and the aggregate ranks for Without DynaMed. Pairs for which there was no difference were excluded from the analysis.
Answered more questions (n = 46)	23 (50)	13 (28.3)	10 (21.7)	.05
Found more answers that changed clinical decision making (n = 46)	25 (54.3)	13 (28.3)	8 (17.4)	.01
Had better overall impact on decision making (n = 46)	28 (60.9)	15 (32.6)	3 (6.5)	.007
Spent less time searching (n = 46)	22 (47.8)	23 (50)	1 (2.2)	.59
Found answers faster (n = 42)	20 (47.6)	22 (52.4)	0	.64
Stopped unsuccessful searches earlier (n = 28)	16 (57.1)	12 (42.7)	0	.69

Table 5.

Results by Per-Question Analysis

Measure	With DynaMed No. (%)	Without DynaMed No. (%)
Total number of questions entered	347	351
Questions answered	263 (75.8)	250 (71.2)
Questions for which the answer changed decision making	224 (64.6)	209 (59.5)
Questions for which participant did not find an answer when the answer would have changed decision making	68 (19.6)	82 (23.4)
Overall impact on clinical decision making, points	156	127
Median time searching (n = 695 questions), minutes	4.95	4.98
Median time to find answers (n = 510 questions), minutes	4.78	4.89
Median time for unsuccessful searches (n = 185 questions), minutes	5.23	5.1