Final - Clinical Assessment and Prediction (Dawes et al.)

Question 1

What are the 2 phases in decision making?

Answer 1

• data collection (mechanical and judgmental scores)
• data integration and prediction (clinical or statistical)

Question 2

What are 3 myths about determinants of good decision making?

Answer 2

• myth of experience
• myth of more info (in reality, primacy/recency effects mean adding more info in between doesn’t make a difference)
• myth of configurality/patterns (idea that “only humans can see patterns”; untrue, can model using a formula)

Question 3

Generally, statistical prediction improves accuracy (beyond clinician prediction) by about __%

Answer 3

10%

Question 4

What did Meehl’s 1965 meta-analysis find about statistical vs clinical predictions?

Answer 4

• statistical better than clinical in 33/51 studies
• statistical equal w clinical in 17/51
• statistical worse than clinical in 0/51

Question 5

What have more recent studies found about statistical vs clinical predictions?

Answer 5

• statistical better vs clinical in 33-46% of studies
• clinical better vs statistical in 6-16% of studies

Question 6

What are 3 reasons why clinicians aren’t better at prediction?

Answer 6

• overweigh successful predictions & forget incorrect predictions
• do not think ab base rates
• do not use all available info consistently or appropriately

Question 7

What are the 2 errors clinicians make in prediction?

Answer 7

• error 1: overweigh positive instances (fail to consider false positives and base rates)
• error 2: similarity heuristic (ignores base rates and validity of test info)

Question 8

When can predictions still be correct despite overweight of positive instances (error 1)?

Answer 8

• if guess is consistent a base rate and base rate is high

Question 9

When can predictions still be correct despite the similarity heuristic (error 2)?

Answer 9

• if base rate is high and test info is valid

Question 10

What are 3 ways to improve clinical judgement?

Answer 10

• systematically consider alternatives
• collect feedback ab decisions/predictions
• think ab statistical prediction issues

Question 11

What CAN clinicians do well in terms of prediction (3)?

Answer 11

• provide input into statistical model (pick what variables are valid and should be included in model)
• generate hypotheses
• provide prediction when no formula exists