Medical Decision Making

Question 0

Deductive Reasoning

Answer 0

–Hypothesis testing
–Comparing current case against a pattern
–Does the patient have pulmonary embolus?

Question 1

Diagnosis

Answer 1

A process and its conclusions

Physicians think about diagnosis in complex ways

Aspect of medicine least well understood by the general public
–“You told me I had exercise-induced asthma and treated me for it. Now you’re saying it’s something different. How can that be?”

Question 2

Inductive Reasoning

Answer 2

–Differential diagnosis
–Find a pattern
–What does the patient have?

Question 3

Diagnostic Process

Answer 3

• Deduction
• Induction

•Pattern Recognition
–Instance based
–Prototype

•Combination

Circular process
•History and physical
•Differential diagnosis
•Diagnostic testing

Question 4

Evaluation of a New Diagnostic Test

Answer 4

•In what kind of patients?
–Among the type of patients who would actually receive the test in clinical practice

•Compared to what?
–An appropriate, reproducible gold standard

•Who should have the gold standard test?
–All the patients being studied regardless of the result of the test being evaluated

Question 5

2x2 Contingency Table

Answer 5

.

Question 6

Prevalence = 10%
Sensitivity = 90%
Specificity = 95%
Total # people = 500

PPV = ?
NPV = ?

Answer 6

45 23 68
5 427 432
50 450 500

Question 7

Prevalence = 20%
Sens = 90%
Spec = 95%
Total people = 500

Answer 7

D+. D-
T+. 90. 20. 110
T- 10. 380. 390
100. 400 500

PPV = 90/110 = 82%
NPV = 380/390 = 97%

Question 8

Influence of Prevalence on Diagnostic Tests

Answer 8

No effect on sensitivity or specificity

Affects predictive values

• Higher prevalence: ^ PPV, v NPV
• Lower prevalence: v PPV, ^ NPV

Question 9

Likelihood Ratios

Answer 9

Ratio of 2 ratios:
Ppl with a given result among ppl with disease / Ppl with the same result among ppl without disease

LR:
(test result / disease positive) / (test result / disease negative)

Question 10

D+     D-
High.                 40.    10
Low.                  100.  150
Indeterminate.   70.    630
                          210.    790

LR high
LR low
LR intermediate

Answer 10

•LRHigh =
(40/210) / (10/790) = 15 x more likely to have disease that ppl with low or indeterminate

•LRLow =
(100/210) / (150/790) = 2.5

•LRInd =
(70/210) / (630/790) = 0.42

Question 11

Prevalence=10%
Sens =  90%
Spec =  95%
PPV = 67%
NPV = 99%

     D+     D- T+    45.    22.    67 T-     5.      428.   433
    50.    450.    500

LRpos
LRneg

Answer 11

LRpos = (45/50)/(22/450)
= 18.4

LRneg = (5/50)/(428/450)
= 0.11

Question 12

Comparison of Predictive Values and Likelihood Ratios

Answer 12

• Both pertain to what does it mean if my patient has a given test result
• Predictive values depend on prevalence of disease

•Likelihood ratios do not
–Derived from sensitivity and specificity which are prevalence invariant

Question 13

Likelihood Ratios and Sensitivity/Specificity

Answer 13

D+ D-
T+ a. b. a+b
T- c. d. c+d
a+c. b+d

•LR(+) =
(a/(a+c))/(b/(b+d))
or
sensitivity/(1-specificity)

•LR(-) =
(c/(a+c))/(d/(b+d))
or
(1-sensitivity)/specificity

Question 14

Bayes Theorem

Answer 14

•General form:
Prior odds hypothesis x Bayes Factor = Final (Posterior) odds of hypothesis

•Form for diagnostic tests:
Pre-test odds of disease x likelihood ratio = Post-test odds of disease

Question 15

Odds and Probabilities

Answer 15

•Odds = probability/(1 – probability)
–p=0.5 odds = 0.5/(1-0.5) = 1:1
–p=0.33 odds = 0.33/(1-0.33) = 1:2 (or 0.5:1)

•Probability = odds in favor/total odds
= odds in favor/(odds in favor + odds against)
–Odds=3:1 p = 3/(3+1) = 0.75
–Odds=1:4 p = 1/(1+4) = 0.2

Question 16

Application of Bayes Theorem
•58 year old man with long smoking history, blood-streaked sputum, lung nodule on CXR
–Pre lung biopsy probability of lung cancer = 33%
•Possible biopsy results: “high”, “low”, “indeterminate”
–LR (high) = 10
–LR (low) = 0.2
–LR (ind) = 1.0
•What’s the post-test probability for a given result?

Answer 16

•Convert pre-test probability to pre-test odds
–p=0.33 odds = 0.33/(1-0.33) = 0.5:1 (or 1:2)
•Multiply pre-test odds with likelihood ratio of result obtained to obtain post-test odds
–High: 0.5x10= 5
–Low: 0.5x0.2 = 0.1
–Indeterminate: 0.5x1=0.5
•Convert post-test odd to probability
–High: 5/(5+1)=0.83
–Low: 0.1/(0.1+1)=0.09
–Indeterminate: 0.5/(0.5+1)=0.33