Behavioral / Stats Flashcards
What is a cross sectional study?
- Purely observational
- Asks, what is going on at this point in time
- Snap shot of exposures and outcomes simultaneously
- Shows risk association with disease
DOES NOT ESTABLISH CAUSALITY
*Disease prevalence is characteristic measure
What CAN’T a cross sectional study establish?
Causality
What is disease prevalence characteristic measure of?
Cross sectional study
What is prevalence?
Portion of people in population with disease at any given time
Weakness of cross sectional study?
Shows correlation, but DOES NOT prove causation
*For example, maybe there is another factor that is causing both the disease and the risk factor you are studying
What is a case control study?
Observation and RETROSPECTIVE study asking, "what happened?" Case: subjects with disease Control: subjects without disease - Previous exposure is being analyzed *Metric is the odds ratio
What is the odds ratio the metric in?
Case control study
What is odds ratio?
- Odds of exposure to risk factor in case (disease) : odds in control (healthy)
- **Does exposure increase odds you have disease
- Used in case control study measuring:
- ***If odds ratio is 9, you’re 9 times more likely to develop disease if you had exposure
Define odds?
Odds = P / (1 - P)
- P = probability event happening
- 1 - P = probability not happening
What is a cohort study?
Observational prospective study
“Does exposure increase risk of disease?”
- If high cholesterol, will you have higher risk of MI
*Relative risk is characteristic measure
- Focusing on if risk is causing disease
What study uses related risk?
Cohort study
What is relative risk?
P1/P2
P1 = risk of disease if exposed
P2 = risk of disease is unexposed
What makes a clinical trial high quality?
- Randomization
- Controlled
- At least double blind
What is single, double, triple single?
Single: patient doesn’t know if getting treatment
Double: Neither doctor or patient know who got drug
Triple: same as above but researcher also does not know
What is phase 1 study?
- Drug studied in 10 - 20 healthy volunteers
- Assess safety, toxicity, kinetics
What is phase 2 study?
- 100s of people WITH disease to assess efficacy
What is phase III study?
- 1000s of people with disease
- Treatment assessed vs. placebo or best available treatment
Trick to remember # of people in phases of study?
The number of the phase, corresponds to number of zeros in patients studied
What is phase IV?
- All people on drug post release to assess rare or LT effects
Questions being asked in 4 phases?
- Is it safe
- Does it work
- Is it more effective
- Can we keep on using it
When drawing 4x4 box what is on left side and top of it?
Left: test
Top: disease
***Just read it left to right, “we are testing for disease”
Test comes before disease as it does on chart
Define sensitivity?
Ability to detect disease when it is actually present
Equation for sensitivity?
TP / (TP + FN)
What are high sensitivity tests good for?
"SNOUT" SN = sensitivity Out = ruling out Ruling things out ***This is true because high value is approaching one with means false negative = zero
Define specificity?
- Ability to indicate non disease when disease is not present
Equation for specificity?
TN / (TN + FP)
What are tests with high specificity good for?
“SPin”
SP = Specificity
IN = Ruling in a disease
*As value approaches 1 it will have very few false positives
*Good for confirmatory test after positive screening
Equation for positive predictive value?
TP / (TP + FP )
Equation for negative predictive value?
TN ( TN + FN )
Which values vary with prevalence of disease?
PPV and NPV
*Sensitivity / specificity DO NOT
Impact of high prevalence of disease on measurements?
Higher PPV, lower NPV
Difference between prevalence and incidence?
Prevalence includes old cases where incidence does not
How does the 4 x 4 chart switch when using it for odd ratio?
Exposure is on left, disease is on top:
- Makes sense in reading left to right as you must first be exposed to get the disease
- *Then layer in A, B, C, D alphabetically as if you were reading
What is the odds ratio equation?
( AD ) / ( BC )
Does odds ratio imply causation?
No, only correlation
- Remember there could be another factor causing both the risk factor and the disease
Relative risk equation?
( A / [A + B] ) / ( C / [C+D] )
Relative risk reduction equation?
1 - RR
What is attributable risk?
Difference in risk between exposed and unexposed group?
Attributable risk equation?
(A/ [A+ B] ) - ( C / {C+D} )
What is absolute risk reduction?
Difference in risk attributable to intervention
Equation for absolute risk reduction?
( C / {C+D} ) - (A/ [A+ B] )
What is number needed to treat?
Number of patients needed to be treated for 1 patient to benefit
Equation for number needed to treat?
1 / ARR (Absolute risk reduction)
What is number needed to harm?
Number of patients needed to be exposed to risk factor for 1 patient to be harmed?
Equation for number needed to harm?
1 / AR
- Think AR = AR assault rifle
What is precision?
Consistency and reproducibility of a test
***AKA reliability
Another name for reliability of test?
Precision
What is accuracy? Another name for accuracy?
How close test measures are to true values
- AKA: validity
Another name for validity?
Accuracy
What does random error do?
Decreases precision and reproducibility of test
What is systematic error?
Reduces accuracy by consistently skewing results in a particular direction
What is bias?
1 result / outcome is systematically favored over another
What is berkson’s bias?
Only people from hospital were selected for study: form of selection bias as they are not as healthy as the general population
- Same idea occuring if you select people healthier than normal
What type of bias is it if certain type of people drop out of study?
Selection
What is procedure bias?
Subjects in different groups are not treated the same
What is lead time bias?
Detecting disease early being confused with better survival
What is the median?
Middle number in set when organized numerically
Measurement most affected by outliers? Least?
Most: Mean
Least: Mode
How is median calculated with even list of numbers?
Average of middle two numbers
What happens to SD in very precise test?
It is decreased
%s falling in first 3 standard deviations?
1st: 68
2nd: 95
3rd: 99.7
Equation for standard error of mean?
Standard deviation / (Square root sample size)
* As sample size increase, SEM decreases
Does standard deviation impact sample size?
No, only impact standard error of mean
Break out of measurements in positive /negative skewed distribution?
Positive: Mode Median > Mean
Best number in skewed distribution?
Median
What is the null hypothesis?
That there is no statistical difference between 2 groups
What is on 4 x 4 box in hypothesis testing?
Results on left, population on top
What is type I error?
Stating there is an effect/difference when one does not actually exist
- Null rejected, experimental accidently accepted
- This is in box B