Pt 725 Ebp Diagnosis

Front 1

How does a diagnosis help us?

Back 1

It helps by limiting uncertainty (it's all about probability)

Front 2

What is pre-test probability?

Back 2

A baseline probability of a certain condition pre-testing (for any given patient)

Front 3

What is post-test probability?

Back 3

The application of a clinical diagnostic test that alters the baseline probability

Front 4

What are the 3 types of diagnoses we'll see as PTs?

Back 4

1) Medical diagnosis (e.g. herniated disc L4-L5)
2) OT diagnosis (e.g. elevated fall risk, dressing apraxia)
3) PT diagnosis (R sided radiculopathy centralizing with repeated extension)

*Want to make TREATMENT BASED diagnoses

Front 5

What are the basic steps in the Diagnostic Process?

Back 5

1) Generate possibilities and their relative likelihood or probabilities
2) Gather new info to clarify your initial diagnostic possibilities
3) Revise pretest and posttest probabilities

*Lead to the concept of ruling in/out a disorder
*Is probabilistic in nature (probabilities are always changing, so has a measure of uncertainty)

Front 6

Why are test and treatment thresholds important in the diagnostic process?

Back 6

Because it tells us the probability of the diagnosis being correct.

-If below test threshold, no testing is warranted (low probability)
-If between test and treatment threshold, further testing is required
-If above treatment threshold, testing is completed and treatment commences (high probability)

Front 7

What is the "index test"?

Back 7

The clinical test of interest (the results of this test go into the 4x4 square to be analyzed)

Front 8

What is the result (in the 2x2 table) if the clinical test is positive and the condition is present?

Back 8

True positive

Front 9

What is the result (in the 2x2 table) if the clinical test is positive and the condition is absent?

Back 9

False positive

Front 10

What is the result (in the 2x2 table) if the clinical test is negative and the condition is present?

Back 10

False negative

Front 11

What is the result (in the 2x2 table) if the clinical test is negative and the condition is absent?

Back 11

True negative

Front 12

How can you tell if a diagnostic test is independent?

Back 12

The proposed test must not be a component of the gold standard test

E.g. Can't use single leg stance and Berg Balance as 2 different tests to measure balance because SLS is part of BBS

Front 13

How can you tell if a diagnostic test study is blind?

Back 13

When the measurers of one test aren't aware of the results of the other test

Front 14

How do you know if you have an appropriate spectrum of patients in your study?

Back 14

1) Similar variety of patients like those you would see in your practice
2) Appropriate array of severity and even some subjects without target condition (avoid "insufficient diagnostic dilemma" = healthy individuals being compared to those with severe cases)
3) Avoid spectrum bias - having an unrepresentative patient selection (lack of sufficient heterogeneity)

Front 15

Should all patients receive the gold standard test?

Back 15

Yes, otherwise study will distort properties of the proposed diagnostic test (workup/verification bias)

Front 16

What is workup/verification bias?

Back 16

When you subject everyone to specific tests but then only those who were reported positive continue on in the study or when the results are verified by different reference standards

E.g. patients with suspected CAD whose exercise test results are positive may be more likely to undergo coronary angiography (reference standard) than those whose exercise test results are negative

Front 17

Do you want to see the proposed test validated in a second group?

Back 17

Yes (but only if validated in a second, INDEPENDENT, group of patients)

Otherwise, results are only considered tentative and preliminary

Especially important with CPRs or diagnostic test clusters

Front 18

Sensitivity is based on what column of the 2x2 table?

Back 18

Left

Front 19

Specificity is based on what column of the 2x2 table?

Back 19

Right

Front 20

The higher the subject number, the _______ the CI.

Back 20

Smaller

Front 21

Tests with high sensitivity have few _______

Back 21

False negatives

Front 22

Does sensitivity rule in or out? Is it better for diagnosing or screening?

Back 22

Rules OUT a condition - better used as a screening tool

Front 23

What does the sensitivity value tell us?

Back 23

It gives us the proportion of patients WITH the condition who have a POSITIVE test result

Front 24

Tests with high specificity have few ______

Back 24

False positives

Front 25

What does the specificity value tell us?

Back 25

It gives us the proportion of patients WITHOUT the condition who have a NEGATIVE test result

Front 26

Does specificity rule in or out? Is it better for diagnosing or screening?

Back 26

Rules IN a condition - better used as a diagnostic tool

Front 27

What are likelihood ratios used for?

Back 27

They're used to reduce the uncertainty about a patient's likelihood of HAVING a target condition

Front 28

How do likelihood ratios help give us a post-test probability?

Back 28

The pre-test probability of having the condition combined with LR gives us a post-test probability

Front 29

If a diagnostic test is POSITIVE which LR do you use?

Back 29

PLR

Front 30

If a diagnostic tool is NEGATIVE which LR do you use?

Back 30

NLR

Front 31

What does it mean if the 95% CI for either LR includes 1.0 (null value)?

Back 31

It means the LR is NOT statistically significant

Front 32

Is a useful PLR always greater or less than 1.0?

Back 32

Greater than 1.0

Front 33

Is a useful NLR always greater or less than 1.0?

Back 33

Less than 1.0

Front 34

You may have an intuitive sense of the pre-test probability based on what 2 factors?

Back 34

1) Patient demographics (and the nature of their complaint)
2) Clinical experience

Front 35

How do you calculate the prevalence of a condition?

Back 35

Take the total of the left column of the 2x2 table and divide it by the total number of subjects

E.g. 200 people in the study, 100 had the target condition, prevalence = 50%

Front 36

What is the point estimate?

Back 36

It is a single value computed from the study subjects giving us our "best estimate" of the true value for sensitivity or specificity in the target population

E.g. Sensitivity = 75%

Front 37

What is the confidence interval?

Back 37

It is a range of values within which we can expect the true value in the population to lie (with a 95% level of confidence, we can conclude that the true value in the target population will be somewhere in this range)

E.g. Sensitivity = 75% (95%CI 70-80%)

Front 38

When is the result of a study statistically significant - based on Sp & Sn?

Back 38

When the CI EXCLUDES the null value (50% is always the null value for Sn & Sp because this is the level of diagnostic accuracy with a coin-flip)

Front 39

When is the result of a study sufficiently accurate for clinical use (clinically significant) - based on either Sp & Sn or LRs?

Back 39

When the point estimate is above (or below with NLR) an appointed level (not necessarily 80% or 0.2/5.0...)

Front 40

When is the result of a study statistically significant - based on LRs?

Back 40

When the CI EXCLUDES the null (always 1.0 for LRs because this is the magnitude of a LR that produces zero shift from pre- to post-test probability)

Front 41

What is the best method for determining ideal cut-scores?

Back 41

ROC curve analysis

It's how you determine that a test score (e.g. BBS 45 or greater) means you'll be at a greater risk for specific symptoms (e.g. higher fall risk)

Front 42

What does the ROC curve plot?

Back 42

It plots the probability of true positives (sensitivity) against the probability of false positives (specificity) for all the possible cut scores

Front 43

How do you select a cut-score from the ROC curve?

Back 43

The best cut score is that score on the continuous scale that yields the highest true positive rate while still minimizing the false positive rate

The closer you are to the upper left border, the higher your specificity and sensitivity values ("break point" in ROC curve)

Front 44

How is a ROC curve analyzed?

Back 44

-Accuracy is measured by the area under the curve (1 = perfect test; 0.5 = worthless test)

-It's a rough guide for accuracy of a diagnostic test

Front 45

What are clinical prediction rules?

Back 45

A statistical approach to identify a parsimonious subset of tests

-Have the greatest predictive power with the least redundancy

Front 46

What is Homan's Sign? Is it clinically useful?

Back 46

A provocative test in order to determine whether the patient has a DVT or not

NOT clinically useful! (>50% false positives, 44-92% false negatives!)

Front 47

What should we (as PTs) do if we think a patient might have a DVT?

Back 47

1) Have them fill out DVT CDR Score card
2) Call physician! Especially if they have 3 or more findings (75% probability of DVT)

Front 48

How does the use of DVT CDR help us as PTs?

Back 48

1) Raises awareness of risk factors for potentially life-threatening DVT
2) Guides urgency of referral to physician