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;
93 Cards in this Set
- Front
- Back
define epidemiology
|
atudy of occurence and distribution of diseases in POPULATION
|
|
clinical approach vs publich health approach
epi is a metho of which one? |
clinical- individual level
public health- studies disease at population epi is method of public health-pop. |
|
epi's 3 factors
|
host
agent environment |
|
define environment
|
all other factors inhibit OR promote disease transmiss.
|
|
agents
|
obvious ones, then u got genetic traits, chemicals, radiation
|
|
some of the host factors
|
personal behaviors
immunization status physicologic states(preg) puberty |
|
environmental factors
|
physical-weather, geology
biological-food, water, air social + cultural- crowding, war |
|
what did epidimiologist start out focusing on?
what are they focused on now |
in past- infectous diseases
now-chronic diseases |
|
infectous diseases are characterized by
|
single, known identifiable cause
|
|
pharmaco-epi studies what
|
drug use patterns, drug use effects, and adverse drug effects
|
|
whats whats diff in adverse event and adverse drug reaction
|
adv event- error while pt is taking med, but doesn't necess. mean it caused event
drug reaction- outcome thats harmful or unpleasent-theres a casual link to drug |
|
define side effect
medication error |
S.E- dose dep effect predicatble + maybe desirable or undesirable
Med Errrr-any preventable event may cause harm to pt |
|
FDA uses what to approve new drug
phase1- 2- 3- |
Clinical trials
1-determines safety 2-small group-for intended pts, tests efficacy 3-thousands of pts, RCT, 1-4 yrs |
|
limitations of clinical trials
|
only studies on hundreds/thousnads of ppl-rare SE will show later
can't see long term effex decay effect efficacy vs effectivness |
|
whats decay effect?
efficacy vs effectivness? |
decay-overtime effect decre. -trial can't show this
efficacy- intended effect doesn't necessarily have effectivness (out of study ppl will have compliance issues) 15 tabs/day |
|
define postmarketing surveillance
|
needed to see long term effects of drugs
means of gathering data about product after approval |
|
T or F
some drugs need to do phase 4 clinical trials AFTER approval |
True
|
|
examples of pharmacovigilance
|
medwatch-info on serious SE reported by HCP
VAERS- vaccine A.E. |
|
define pharmacovigilance
|
science of detecting, assessing, and preventing A.E. of meds
|
|
whats ultimate goal in establishing relationships between drugs and outcomes
|
establishin a casual relationhip (casuality)
|
|
whats criteria for causality
|
1) correlation-consist pattern of change
2) temporal ordering 3)rule out alternative explanations |
|
whats internal validity
|
approx truth of inferences for cause-effect relationships
|
|
what kind of variable is independent
|
predictor- its variable being manipulated durr
|
|
what kind of varibale is dependent
|
outcome -whatever result of changing indep durr again
|
|
define confounding
|
extraneous varibale that correlates with indep and dep variable (unforseen correlation factor)
need to control for this |
|
type 1 error
|
false positive
null is true but reject null |
|
type 2
|
false negative
null is false but accept null |
|
define sampling
|
selecting units(ppl) from population so you can later generalize results back to population
|
|
random selection
|
different units are selected by chance
|
|
parameter vs statistics
|
para-mean if you were able to sample entire POPULATION
statistic- value estimated from sample data |
|
obsv vs experiment study
|
obsv-no intervention or exp
exp- manipulation of factor(exposure) and randomization of ppl |
|
in regards to data collection
define primary define secondary give examples |
primary-investigator is 1st to collect info
-medical exams, interviews, observations seondary- data collected by OTHERS of other purposes than study -medical/employ records, census data |
|
define a study design
|
translate conceptual hypothesis into operational one
|
|
goal of study may be? (3)
|
exploratory- when knowledge is poor, small scale
descriptive-from hypothesis, ex-cross sect, case study analytical-test hypotheses, ex case control, RCT, cross sect. |
|
spectrum for causality
|
estab. cas-----Gen hypothesis
RCT>Cohort>casecontrol>cross sect. |
|
observational design-
types |
alt to exp design, good for drug use patten+ outcome-S.E.
1)case report-clinical exp of 1 pt on drug 2)case series- same but mult. pts |
|
case series are of value cuz they
|
study cases of similar disease
and give clinical education, audit and resarch value |
|
give an example of case series mentioned in class
|
HIV epidemic
dr's saw pts with kaposi's sarcoma- but not in normal pts(mediaterrian men with cancer) |
|
whats an ecologic study
whats missing in this design |
utilizes aggregate data or combine data with individual data
missin relationship btwn exposure and outcome at individual level(incomplete design) study must be interpreted carefully to avoid eco fallacy |
|
define ecological fallacy
|
taking aggregate data and applying to specific individuals
CAD higher in richer cities |
|
cross sectional study
|
disregards length of time (done at 1 pt in time)
provide estimates of prevalence of condition but not incidence useful in developing new hypothesis not causuality(cause inabillity to determine if exposure was before outcome) |
|
case-control study
whens it useful? |
useful in rare outcomes
study sample of ppl with cancer and group without outcome of interest as control temporal order of exp + outcome is paramount** |
|
whats a common element used in case-control to avoid possible confounding
drawbacks of case-control |
matching
misclassifaction selection bias recall bias |
|
longitudinal study
|
2 or MORE observations are collected for every unit in study
offer ability to see changes in study unity and across units |
|
2 studies used widely in epidemiology
|
case-control
cohort |
|
whats common application of cohort
|
used when its unethical or impossible to intentionally expose pt to drug or intervention
|
|
cohort drawbacks
viewed as |
exp+time consuming
not for rare outcomes as modified approach to RCT-but no random assign |
|
TEMPLES neighbor OSMAR isnt trying to find job, TEMPLES thinks all hispanics are lazy (true story)
is this ecological fallacy? |
NO
individual->group = stereotype group->individual= ecolog fallacy |
|
what are traditional epidemiologic studies
|
observational- no intervention or tx
|
|
Random assigment?
yes- no- |
if yes-experiemntal
if no, is there control grp or multiple measures? if yes-quasi exper. no- non-exp(observ.) |
|
experimental designs are G.S. in terms of establishing...
|
internal validity
|
|
random selection vs random assignment
|
rand sel- collecting ppl from population to sample (external validity)
rand assign- 2 grps will preform same, any diff is cuz chance determines internal validity |
|
experimental design classified into
|
signal to noise approach
signal-related to key variable tying to measure noise-all factors(confounders) make it hard to see signal want signal to be higher than noise |
|
T or F
increasing the signal is better than decreasing the noise |
F
they both increase the quality of research |
|
signal-enhancing designs are called...
whats prototypical factorial design |
factorial designs
RCT |
|
what are 2 major types of noise-reducing designs
|
covariance and blocking designs
|
|
for assignment to group, what does:
R, N,C stand for |
R-random
N-nonequal grps C-cutoff |
|
is everyone in a factorial design (2x2) getting same tx?
|
No
some get X11, others get X22 |
|
for a 2x2..how many factors are there?
for a 2x2x3 how many factors? |
2 factors(prolly dose + duration)
3 factors |
|
how many levels are there in 2x2?
2x2x3? how many groups? |
F1-2, F2-2= 4 groups
F1-2, F2-2, F3-3= 12 groups |
|
if we are given a pre-intervention test and score avg is 90....what is the score once the pts are broken up into groups
|
about same approx 90
|
|
what is the null effect
|
its situation where diferent tx combos have NO effect
all scores for tx and setting =90 |
|
for a factorial design of 3x3x2 how many effects will there be
|
there will be 3 effects,
there is 1 effect for every factor |
|
whats a main effect
vs a interaction effect |
main- you look at 1 factor at time + # in particular lvl is lower than any number(regardless of lvl) it has main interac on outcome variable
if no level with the lowest # the factor has interaction effect on outcome |
|
RCT most commonly studies info abou...
|
ADV. DRUG RXS and EFFICACY or EFFECTIVNESS of healthcare services or health technologies(med dev, surgery)
|
|
randomization ensures equal distibution of
|
confounders
|
|
main goal of randomized block design
what kind of strategy is this? |
to reduce noise or variance in data
analytical- it doesnt affect anything with research participants, it groups ppl in data analysis |
|
you can't do a block design unless
|
you collected data on potenetial confounders
|
|
what seperates regression discountinuity (quasi experimentl design) apart from other prepost designs
|
method of participants are assigned to condidtions
assigned based on cutoff score or preprogram measure |
|
what are steps in doing a regression discontinuity design
|
1- measure QOL for everyone pre and post w/o txing
2-set cutoff-tx ppl <50QOL, and no tx >50QOL 3- see if results <50QOL line moved up or down compared to the line for >50QOL |
|
what is the difference between the cutoff groups regression lines
|
its the discontinuity...its in the regression lines at the cutoff point
|
|
what is a maturation threat
how to correct |
something that changes naturally (ex-kids taking math class in school+ tutoring)
is tutoring helping or is it really learning over time form school add a control group R---O---X---O (orig study) ------------------------------------ R---O---X---O tutoring grp R---O--------O non tutoring |
|
how do you address an instrumentation threat
|
R---O---X---O
R---O---O---X---O add another observation ex-kids learn prof always has B as answer do new study with both exams being B, then change last exam randomly |
|
correlation indicates
its purpose |
degree of association btwn 2 variables, "r" but not if one caused the other
to make a prediction about 1 variable based on known v. |
|
what are 2 correlation directions
|
positive- x incr. y incr
negative xin and y dec. education + prejudice |
|
positive r
negative 1 is a negative # strong or weak? |
0-1
-1-0 doesn't indicate strength, just look at number (-.8>-.2) |
|
does correlation establish causation?
|
No
need temporal order, correlation, and rule out alt |
|
pearsons correlation coefficient (r) determines
|
stregnth and direction btwn x and y, measured at INTERVAL LEVLS
|
|
steps to do a correlation problem
|
1) calc r..make sure denom is done by summing (x-x1)^2 each the multiply total by total y^2
2) calc t value 3) compare t value to table, if > then its statis signif. |
|
what are requirements for using pearsons r
|
straight line
interval data random sample normally distributed x + y |
|
simple regression analysis
|
like correlation, but we are interested in seeing if changes in x CAUSES change in y
|
|
what is R^2
|
coeffiecent of determination
its % of variance x(dep) accounted by variablitiy in y(indep) |
|
what are residuals
|
HINT** its not the left over cum on your moms mouth
diff btwn predicted values and actual value of dependent variable (y) errors we can't predict |
|
error =
what does a + error mean |
Y(actual value) - y^(predicted)
+ means y^ is under predicting - is over predict compared to Y |
|
if you don't know x
how do you determine error |
_
Y(actual)- y (avg) |
|
if you have x
how do you determine error |
y(actual)- y^(predicted)
|
|
how do you determine SStotal
|
you do this when don't know x
its the numerator in r or _ E(y-y) ^2 |
|
how do you determine SSerror
|
this is done when you know what x is
E(y-y^ )^2 |
|
what the equation for SSreg
|
SSreg= SStotal - SSerror
|
|
whats equation for Proporionate reduction in error or PRE
|
SSreg/ SStotal
|
|
how do you determine r2
and what does it express |
well can do r squared
or find PRE and then put in percentage it expresses % of variance in y explained by x |
|
how do you test statistical sig of regression coeff (r^2)
|
SSreg/dfreedom (1)
-------------------------------- SSerror/ N-2 this is equal to F, compare this umber to Ftable to determine sig or not |
|
steps in making a simple regressio formula
|
1)use formula to determine b(slope) SP/SSx and intercept a=y(avg)- bx(avg)
2) make new table-use x to determine y^(predicted) 3)do SStotal equat and SSerror (its y- new predicted) 4) find SStotal 5) calc PRE = r2 6) calc F and compare to table to determine signif. |