Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
31 Cards in this Set
- Front
- Back
|
Risk factors
|
factors associated with development of disease (ie- smoking and lung ca)
|
|
|
Prognosis factors
|
factors which predict the outcome of disease (ie - stage of Ca/spread to other organs)
|
|
|
4 Ss
Systems |
1 - Computerized decision support
|
|
|
4 Ss
Synopses |
2 - EBM journal abstracts. ACP journal club (ACP contains studies on prognosis)
|
|
|
4 Ss
Summaries |
3 - Synthesis UptoDate, websites (cancer.gov), Cochran reviews
|
|
|
4 Ss
Studies |
4 - Original published articles
|
|
|
Search strategies for prognosis
|
PubMed: Clinical queries "prognosis" or cohort studies as MESH subheading
ACP Journal club, UpToDate |
|
|
RCT limitations
|
Unethical study
Rare events - requires very large study to show statistical difference = costly and impractical |
|
|
Prospective cohort studies strengths
|
Stronger associations than case control
- no recall/interviewer bias - more control over data collected |
|
|
Prospective cohort studies limitations
|
Confounding
- pt groups not identical - differental exposure (differences besides exposure determine outcome) - does not account for unmeasured variables |
|
|
Causes of confounding (prospective cohort trials)
|
Unrepresentative sample
Subgroups of pts with different prognoses - different stages in disease - subgroups with different prognoses Exposure not randomized -pts/MDs choose exposure Leads to over or underestimation of treatment effect |
|
|
Referral bias
|
pts recruited from specialized settings/systematically different (unrepresentative) from population of interest
|
|
|
Methods to reduce confounding
|
Avoid referral filters
Pts in similar, well defined point in disease progression Subgroups with different risk of outcome considered separately Account for confounding due to choice of exposure by pt or MD Adjustment analysis for baseline differences |
|
|
Correlation coefficient (r)
|
measures the strength of relationship
-1 strongest possible negative relationship 0 no relationship 1 strongest positive relationship |
|
|
Correlation coefficient, strength
|
strength > 0.80 is strong, but also must determine statistical significance to prove that correlation is not due to chance.
|
|
|
Multivariable regression with continuous variables
|
linear regression analysis
|
|
|
Multivariable regression with dichotomous variables (nominal data)
|
requires logistic regression analysis
|
|
|
Multivariate regression
|
determines the contribution of the predictor variables make to the outcome
|
|
|
Adjustment analysis
|
adjusts for different prognostic factors between comparison groups
uses multivariate regression analysis |
|
|
Ratio
Odds |
Event/Total
Event/Non-event |
|
|
Risk ratio
|
Event/total in treatment group divided event/total in control group
|
|
|
Odds ratio
|
event/nonevent in treatment group divided by event/nonevent in control
|
|
|
OR and RR are same when...
|
risk < 0.30 and when OR ~1
(OR=1 means no effect) Compared to RR, OR makes effect size look larger |
|
|
Loss to FU
|
large loss relative to number of outcomes can have great effect on results
|
|
|
When outcome measures subjective judgement (ie - EKG interp, etc)
|
assessors should be blinded to which group the pt is in.
|
|
|
Value of cohort studies
|
-therapy decisions
-conditions with good prognosis -> not to treat -conditions with poor prognosis ->treatment decisions |
|
|
Clinical prediction rules come in handy when
|
1) complex decision making required
2) clinical stakes are high 3) cost savings possible without compromising care |
|
|
Development of Clinical prediction rules
|
1) Derivation and ID of predictive factors
2) Validation 3) Impact analysis |
|
|
Clinical Prediction Rules
1 - Derivation |
ID factors with predictive power
Level of evidence IV |
|
|
Clinical Prediction Rules
2 - Validation |
Narrow - apply to similar pop
Broad - apply to different pop in different setting/multiple settings Level of evidence III/II |
|
|
Clinical Prediction Rules
3 - Impact analysis |
RCT to test if rule changes MD behavior, improves pt outcomes, reduces costs
Level of evidence I (best) |
