This site is 100% ad supported. Please add an exception to adblock for this site.

MED2042 WEEK 9 - Evidence based diagnosis


undefined, object
copy deck
What is the likelihood ratio?
- Adjustment factor
- Used to convert pre-test probability to post-test probablity

- For every sensitivity (<100) there is a specificity that could render the LR=1
- all tests are imperfect
What is the likelihood ratio (LR) for a series of tests?
In this instance the LR of a series of two tests is the product of the LR's of the independent tests:

LR (Test1 and Test 2 in series) = LR Test 1 result * LR test 2 result
What are predictive values?
"This test has a good positive predictivce value"

This statement includes information based on test result (LR) and pre-test probability...

It provides false reassurance:
- demonstrate that PV's change as a function of prevalence (estimate of prior probability) whilst Se and Sp remain stable
How do we assess pre-test probability?
Stratify Pre and Post test probability of PE: Low/Medium/High
- Estimate PTP (Low/Medium/High):
- clinical judgement: empirical assessment based on experience
- use scoring system: decision support

Use post-test probability to decide whether to treat or not
- based on risk of pathological consequences (Low/Medium/High)
How does the prdictive power of a test result depend on the pre-test probability?
As pre-test probability rises:
- likelihood that postive test is a true positive reult rises
- likelikhood that negative test result is a false negative result also rises
- (converse is true as PTP falls)

e.g. Well's scoring system
- based on lit review and expert opinion
- Derived using 80% of subjects in the study then validated on remaining 20%

Wells simplified Clinical probability score for suspected PE:
- clinical signs and symptoms of DVT (3)
- PE probability >= likely as alternative diagnosis (3)
- Immobilisation (>3 days) or Surgery (4 weeks) (1.5)
- Previous DVT (1.5)
- HR > 100 (1.5)
- Hemoptysis (1)
- Active cancer (1)

PTP low <2
PTP moderate 2-6
PTP high >6
What are some important things about evidence-based diagnosis?
- Most tests are imperfect:
> adjust disease probability
> require estimate of PTP
- LR's adjust PTP using both sensitivity and specificity
- Clinical saying (aphorisms) like "affected until proven otherwise..."
> reall mean there is a low threshold for further tests i order to confirm/exclude diagnosis
- "Predictive balue of a test result..."
> interpret this statement with caution
- Specificity is not an intrinsic property of a test...
> depend on the characteristics of the derivation population.
What is something important to remember about specificity?
- tells us about unaffected individuals who have a positive test result
- Depends on characteristics of the test population
- Affected by prevalence of other disorders that may cause positive test
> a test may be positive for >1 condition

It is important to know whether the test has been evaluated in a population where there may be more than one cause of a positive test result
Why is the process of diagnosis complex?
- In patients who are ill we need information about the accuracy of diagnostic test AND local information about the prevalence of disorders in patients like ours
- In population sceening of well people, we need information about whether people benefit from screening
- Case finding in (otherwise) asymptomatic patients needs different evidence again
Why is reliable evidence about diagnostic test accuracy hard to find?
- Methodological quality of studies of diagnostic tests is low
> The Standards for Reporting of Diagnostic Accuracy (STARD) sterring group aims to improve the accuracy and completeness of reporting of studies of diagnostic accuracy. The group describes and explains the development of a checklist and flow diagram for authors of reports
- Biased estimates of diagnostic test accuracy and poorly designed studies are misleading for clinical decision makers
What is the 4S pyramid of research evidence?
SYSTEMS - Use explicit review processes to present high quality pre-appraised evidence
SYNOPSES - secondary publication e.g ACP journal club
SYNTHESES - e.g. cochrane collaboration
STUDIES - original studies; use MEDLINE diagnosis, search filter on PubMed
What might happen in the future with EBM?
Most studies of diagnostic accuracy only compare a test with a reference standard. Is this helpful?

Types of new diagnostic test:
- REPLACEMENT - safer, quicker, more reliable etc
> compare diagnostic accuracy of both tests
> paired study design: Conceal results of each test in half the subjects
- TRIAGE - used before existing test e.g. D-dimer
> Paired study design: Conceal results of each test in half the subjects
> Limited verification
- INCREMENTAL - Add-on: used after existing test
> May be limited to subgroup depending on test result, e.g. when new test is more accurate but expensive e.g. PET scan after CT for cancer staging
> paired study design with limited verification
What questions should be ased for studies of diagnositc accuracy?
1. What is the existing diagnostic pathway?
2. How does the new test compare with the existing test? (e.g. accuracy and other features)
3. What is teh proposed role of the new test? (Replace, Triage, Add-on)
4. Given proposed role: what is the best measure of test performance and how can it be obtained efficiently?

Deck Info