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;
65 Cards in this Set
- Front
- Back
|
Prevalence
|
= total number of case / population at risk
NOT a measure of risk. |
|
|
Incidence
|
= newly diagnosed cases / (population * time)
|
|
|
Prevention: 3 types
|
1 Primary: Prevent occurence
2 Secondary - reduce severity 3 Tertiary improve function following disease |
|
|
Odds ratio
|
# of cases / # of non cases
|
|
|
Incidence rate
|
AKA IR = # of new cases / population year...
USE TO ASSES effectiveness of interventions |
|
|
Attributable risk
|
ASSESS INTERVENTIONS..
AR = IR (exposed) - IR (unexposed) excess risk.. could also use for treatment |
|
|
Types of Data
|
Numerical - quantitative
Categorical - qualitatite.. Also have nominal (ordered), Ordinal (NOT ORDERED ie race) |
|
|
Normal distribution stuff
|
68% of population +-1 SD
95% of population +-2 SD |
|
|
Prevalence (letter formula)
|
Prevalence = (A+C) / (A+B+C+D)
|
|
|
Sensitivity
|
Sensitivity = A / (A+C)
|
|
|
Specificity
|
Specificity = D / (B+D)
|
|
|
Cheap safe treatments want?
|
SENSITIVE TESTS... get all those positives
IE minimize False Negatives.. |
|
|
Painful expensive stressful diagnosis
|
SPECIFIC TESTS... get those negatives.
Why? You want to want to MINIMIZE False positive.. they will undergo treatment :( |
|
|
Predictive Value (+)
|
PPV..I think of these as positive predictive prevalence.. looks like a prevalence
PPV = A / (A+B) |
|
|
Predictive value (-)
|
NPV.. i think of as negative predictive prevalence.. looks like prevalence formula.
NPV = D / (C+D) |
|
|
Predictive values lose accuracy WHEN?
|
when population prevalence is LOW.. this is due to just having so many false positive
|
|
|
Validity?
|
AKA Accuracy.... the likelihood that test will be correct
|
|
|
Precision (reliability) the likelihood
|
(reliability) the likelihood that repeated measurements will have same result..
|
|
|
Series testing
|
Patient must test positive twice.
High sensitivity then high specificity.. Decrease sensitivity, increase specificity |
|
|
Parallel testing
|
Patient can test positive for either of two tests.
Increase sensitivity, decreases specificity |
|
|
Likelihood Ratio
|
LR = Odds of True Pos / Odds of False Pos
OR LR = Sensitivity / 1- Specifity |
|
|
Odds to prevalence conversion
|
Odds = Prevalence / 1 - odds
|
|
|
Study Designs:
Cohort |
Classify by risk factors, measure outcome. Risk factors so you use relative RISK.
RR = (A / (A+B)) / (C / (C+D) Dyanmic cohort - come and go prospective cohort - have no yet occured retrospective cohort - have already occured |
|
|
Advantages of cohorts
|
Incidence and natural history.. you get a temporal look at things.
The only study which shows incidence. avoid survivor bias avoid reporting bias multiple outcomes |
|
|
Disadvantages of cohorts
|
Ineffect for rare disease
confounding may occur Sub-clinical disease may affect risk loss to followup |
|
|
Study Design:
Case Control |
Classify by outcome, examine risk factor.
Outcome is ODDS ratio. Looking at cases so look down the rows. Odds ratio = Odds of cases / Odds of Controls Odds ratio = (A/C) / (B/D) For very rare disease the relative risk is approximate to the odds ratio |
|
|
Case control vs cohort
|
Cohort separate population into those with risk and those without and follow them through time to determine outcome.
Back for rare disease. Case control examine outcomes and compare the levels of exposure in case and control groups. |
|
|
Cross sectional
|
determine disease and exposure at same time
Find prevalence. Survey based. Advantages: Quick and cheap descriptive info examine associations Disadvantages: Temporal associations not clear Selection bias (ie healthy worker) Shows association only not causality Disadvantages |
|
|
Experimental Trial
|
A TYPE OF COHORT!!
Randomized assignment. Outcome cure not disease usually. Randomization asures comparability Prevents biases Maintain group comparability by keeping groups intention to treat |
|
|
Random error
|
P-value the probability of false positive.. possibility that result is due to chance alone.
Reduce by large sample size and better measurement |
|
|
Selection bias
|
Do subjects accurately represent the target population?
|
|
|
Information Bias
|
Are measurements accurate? Do patient overstate coffee drinking if they have mi?
|
|
|
Confounding
|
For it to occur, it MUST be unevenly distributed between groups. Confounding factor associate with bother exposure and outcome. Reduce by:
Matching make sure controls also smoke OR Specification-limit to non smoker Stratify and regression analysis |
|
|
Reduce error by
|
large population, higher participation.
Specify or match sparingly |
|
|
Selection bias types:
|
Lead Time bias - pancreatic
Length time bias - prostate |
|
|
Categorical Data is reported as
|
Frequencies, proportions, or precentages...as bar graph
|
|
|
Means are ... but medians are ...
|
Means are sensitive to outliers, but medians are robust.
|
|
|
Standard deviation is used for...
|
the individual..YOUR patient population/sample
|
|
|
Standard error is used for...
|
The ENTIRE population
|
|
|
A sample may not be bell-shaped, but ..... ALWAYS ARE
|
a distribution of man sample MEANS always are
|
|
|
Standard error =
|
Standard error = SD / sqrt(n)
|
|
|
95% of [...] would be between mean+-2*SE
|
Sample means. This is the 95% cofidence interval
|
|
|
Confidence intervals mean:
95% of all samples would give an interval that includes the... |
true mean! But we dont know which ones :(
|
|
|
Predictions about individuals use mean and the [...]
|
Standard deviation
|
|
|
Null hypothesis says:
|
No difference or relationship. What we see can be explained by chance variation alone
|
|
|
Alternative Hypothesis says
|
There is in fact a relationship or difference. Chance is unlikely to explain
|
|
|
Type 1 error
|
incorrectly reject H0...
Hard to undo! |
|
|
Type 2 error
|
incorrectly accept H0
|
|
|
Guard against Type 1 errors using significance level...also known as
|
Alpha.. the probability of type 1 error we can tolerate
|
|
|
p value?
|
probability that we would get a result this extreme JUST BY CHANCE... if null hypothesis were true.
|
|
|
If p-value less than alpha...
|
we reject null hypothesis
|
|
|
We guard against the probability of type II error by...
|
Large enough sample size!
|
|
|
Probability of type II error is..
|
Beta! power = 1-Beta
|
|
|
Power is...
|
probability that we make the right decision when there is a true difference
|
|
|
Increase power with...
|
larger sample size
significance levels |
|
|
Direct adjustment
|
choose standard population... apply your specific rates to the standard population to compare
|
|
|
Indirect adjustment
|
Take rates FROM reference standard population and apply to ours. This can calculate standardized mortality rate.. see how much higher our mortality is than reference
|
|
|
Limited english proficiency (LEP): Time with physicians
|
Same as english. most physicians believe more LEP, but that is wrong
|
|
|
LEP: More iatrogenic harm.
|
More communication errors, more adverse affects
|
|
|
LEP: satisfaction
|
LEP patients have more dissatisfaction with healthcare
|
|
|
LEP: Poorer access to care, quality of care, and health status
|
LEP strongly impeded access to care, poorer quality of care
|
|
|
LEP children
|
Triple odds of poor health status. double odds of day in bed for illness, greater odds of not being brought in for needed medical care
|
|
|
LGBT- access to healthcare and health insurance
|
Hetero have more health insurance.
LGB adult more likely to delay medical care |
|
|
LGBT- societal bias
|
More LGB have cancer
LGB youth more likely to be threatened or injured by weapon LGB youth more likely to be in physical fight |
|
|
Impact of societal biases on mental health and well being
|
LGB more likely to experience psychological distress in past year
more likely to have suicide ideation youth more likely to attempt suicide |
