Skip to main content

Table 2 Participants, attributes and outcomes of included studies

From: Cognitive biases associated with medical decisions: a systematic review

Author

Type of participants

Number of vignettes or medical cases

Number of attributes

Based on Guidelines

Outcome measure

Type of outcomea

Type of analysis

Data qualityb

Main findings

Redelmeier

GPs and Neurologist

4

10-11

yes

Treatment recommendations

4

unadjusted

5

Multiple options decreased the likelihood of medication prescription for pain and carotid endarterectomy by 26 % and 35 %, respectively

Ross

GPs

3

NA

No

Descriptive

5

adjusted

6

GPs were less likely to arrange a further consultation for female patients than for male patients (OR = 0.55). GPs with a pessimistic belief about depression were less likely to discuss non-physical symptoms or social factors; More experienced GPs were less likely to conduct a physical examination (OR = 0.60).

Graber

GPs

2

8-9

No

Descriptive

1

adjusted

4

GPs were less likely to believe a serious medical condition among patients with history of depression or somatic symptoms

Sorum

GPs

32

5

yes

Probability of ordering a test

4

adjusted

4

PSA were more likely ordered among GPs with discomfort for uncertainty and those who expressed regret.

Baldwin

Pediatric ED physicians

397

NA

No

Admission rates

4

adjusted

5

Risk aversion scores higher for physicians with >15 years of experience. Admissions rates did not differ between high and low risk adverse physicians (31.1 vs 30.1; p = 0.91). Adjusted admission rates did not different between high and low discomfort with uncertainty (32.3 vs 29.7; p = 0.84)

Friedmann

Medical students (72), residents (72), physicians (72)

36 (9)

>20

No

Diagnostic accuracy

5

adjusted

4

Overconfident found in 41 % of residents and in 36 % faculty.

Reyna

GPs and specialists

9

NA

Yes

Diagnostic accuracy and management

6

adjusted

5

Physicians deviated from Guidelines in terms of discharge. GP were more risk averse and less likely to discharge patients. Experts achieved better case-risk discrimination by processing less information

Bytzer

Specialists

5

NA

No

Diagnostic accuracy

6

unadjusted

4

Only 23 % endoscopists gave the same diagnosis for the two identical video-cases. The great majority were affected by prior information bias.

Dibonaventura

Physicians

2

11--12

No

Descriptive

4

unadjusted

4

Naturalness bias present in 40 %, omission bias in 60 % of participants

Mamede

Residents

8

NA

No, confirmed diagnosis

Diagnostic accuracy

5

unadjusted

5

Availability bias increased with years of training. Clinical reasoning ameliorate this bias

Mamade

internal medicine residents (34) and medical students (50)

12

>20

No

Diagnostic accuracy

6

unadjusted

3

Conscious deliberation improved the likelihood of correct diagnosis in physicians, but not in medical students problems were complex, whereas reasoning mode did not matter in simple problems. In contrast, deliberation without attention improved novices’ decisions.

Gupta

ED Physicians

6

>20

No

Descriptive

1

adjusted

6

Outcome bias tends to inflate ratings in the presence of a positive outcome more than it penalizes scenarios with negative ones.

Perneger

GPs and specialists, and patients (1121)

1

5

No

Rating of new drug

6

adjusted

4

Physicians and patients provided higher value to the hypothetical new medication when presented in relative terms. Compared to descriptive information, relative mortality reduction (OR 4.40; 3.05 – 6.34), Number needed to treat (OR 1.79; 1.21 – 2.66), and relative survival extension (OR 4.55; 2.74 – 7.55) had a more positive perception.

Stiegler

Residents (32), Faculty (32)

20

NA

Catalogue of common cases

Management

1

unadjusted

4

1. Developed a cognitive factor/bias catalogue, 2. Top 10 cognitive biases and personality traits: anchoring, availability bias, omission bias, commission bias, premature closure, confirmation bias, framing effect, overconfidence, feedback bias, and sunk cost.

3. Errors perceived by faculty to be important to anesthesiology were indeed observed frequently among trainees in a simulated environment.

Ogdie

Residents

41

NA

No

Descriptive

6

unadjusted

3

Most common biases: anchoring (88 %), availability (76 %), framing effect (56 %), overconfidence (46 %)

Meyer

Physicians

4

6-11

No

Diagnostic accuracy

2

unadjusted

4

Higher confidence was related to decreased requests for additional diagnostic tests (P = .01); higher case difficulty was related to more requests for additional reference materials (P = .01).

Crowley

pathology residents, fellows and staff pathologists

40

NA

No

Diagnostic accuracy

6

unadjusted

4

Overall, biases occurred in 52 % of incorrect cases compared to 21 % correct. Most common biases-Availability (20 %) and satisfying biases (22.5 %) the two most common. All the rest, less than 10 %.

Saposnik

Residents, internists, emergency physicians and Neurologist

10

5-7

No

Probability of death or disability

6

adjusted

5

Higher confidence was not associated with better outcome predictions. 70 % of underestimated the risk of the death or disability, 38 % overestimated death at 30 days.

Msaouel

Residents

2

4, 5

No

Descriptive

1

adjusted

5

Gambler’s fallacy in 46 %, conjunction bias 69 %

Yee

Specialists (Obstetricians)

3488

NA

No

Management

6

adjusted

7

Physicians with a higher tolerance of ambiguity were less likely to deliver patients by operative vaginal delivery (11.8 % vs 16.4 %; p = 0.006). The effect disappeared in the adjusted analysis (OR 0.77, 95 % CI 0.53-1.1)

  1. NA not available, GP general practitioners
  2. aType of outcome measured: 1 = probability, 2 = rating, 3 = ranking, 4 = yes/no choice, 5 = discrete choice, 6 = more than 2 alternatives
  3. bData quality assessed by the Newcastle-Ottawa Score. See details in the text and Additional file 2