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;
61 Cards in this Set
- Front
- Back
What are examples of continuous variables?
|
1) blood pressure
2) Blood glucose level *a test for which there is no +/- result. |
|
What is the trade-off btwn sensitivity and specificity?
|
If you increase the sensitivity (by lowering the cutoff level) you decrease the specificity; if you increase the specificity (by raising cutoff level) you decrease the sensitivity
|
|
Define a bimodal curve
|
The type of dist. in which there are 2 peaks
permits separation btwn idividuals who had no prior exp. w/ dz X and those who had exp. W/ dz X Ex: ppd test for TB |
|
define a unimodal curve:
|
a single peak distribution
In order to separate those w/ dz from those w/out dz a cutoff is necessary |
|
Determining cutoff levels are very much related to
|
the level of importance place on FP and FN for the dz in question
|
|
for a two stage (sequential) testing to find the net sensitivity
|
# of TP (correctly id) in 2nd test
------------------------------ Total # of people w/ dz in the 1st test |
|
for a two stage (sequential) testing to find the net specificity
|
The combined TN in test 1 and 2
------------------------------ Total # of people w/out dz in the first test |
|
How can you handle confounding in designing and carrying out the study?
|
1.individual matching
2. group matching |
|
How do you handle confounding in the analysis of data?
|
1. stratification
2. adjustment |
|
What is the PPV(postive predictive Value)?
|
in clinical setting it is the probability that the pop. that tested + has the dz
|
|
What is the NPV(Negative predictive value)?
|
the probability that the pop. that tested - does not have the dz
|
|
What is the formula for:
PPV= |
TP/TP+FP
|
|
What is the formula for:
NPV= |
TN/TN+FN
|
|
What 2 factors affect the PPV?
|
1) the prevalenc of a dz
2) When its a rare dz the specificity of the test being used |
|
What is the relationship btwn Predictive value and Prevalence?
|
they are directly related if you have an increase in PV there is an increase in Prev.
|
|
What is the relationship btwn Predictive value and specificity?why is this so?
|
An increase in Specificity has a much greater effect on Predictive value than an increase in sensitivity because of the infrequency of dz.
|
|
intrasubject variation
|
variation within individual subjects; taking into consideration under which the test were performed
|
|
interobserver variation
|
The ability to express the extent of agreement in quantitative terms
|
|
intraobserver variation
|
when the test and examinations differ in the degree to which subjective factors enter the observers conclusions
|
|
How do you calculate the overall percent agreement?
|
add the #'s of all the cells(cross tabulation) inwhich reading by both observers agreed
------------------------ x 100 Total # of readings by combined observers |
|
Percent agreement when using paired observation btwn obs.1 and obs. 2
|
a+d
-------- a+b+c+d |
|
What can you do when the D cell (neg. observation pop.) is large
|
when cell d is large let d=disregard and exclude it from the percent agreement formula:
a ----- x 100 a+b+c |
|
What is the validity of a test?
|
Its ability to distinguish btwn who has dz and who doesnt
|
|
sensitivity
|
The ability to identify correctly who has dz
|
|
specificity
|
The ability to identify correctly who does not have dz
|
|
true positives (TP)
|
people w/ the dz who were correctly called postive by the test
|
|
true negatives(TN)
|
people w/ the dz who were correctly call negative by the test
|
|
False Positive (FP)
|
people who dont have the dz who are erroneously called postive by the test
|
|
False Negative (FN)
|
people who do have the dz who are erroneously call negative by the test
|
|
what are important issues with false postives?
|
-the burden on healthcare
-emotional effect on people -delabeling is never fully achieved |
|
what are important issues with True negatives?
|
-if the dz is serious it could be a death sentence
-depends on the nature and severity of the dz being screended for. |
|
What is the definition of a bias?
|
"any systematic error in the design, conduct or analysis of a study that results in a mistaken estimate of an exposure's effect on the risk of disease."
|
|
selection bias
|
observe an assoc. btwn exposures and outcome that arises from the process of identifying the study pop.
|
|
exclusion bias
|
when a different set of criteeria are used for the cases and controls
|
|
to be a confounder in the relationship btwn exposure and a dz the factor must be:
|
A risk factor for the dz
assoc. with the exposure but not caused by the exposure |
|
confounding
|
confounding is a valid finding that describes the nature of the relationship among several factors and the risk of disease. However, failure to take confounding into account in interpreting the results of a study is indeed an error in the conduct of the study and can bias the conclusions of the study.*confounding is a function of the complex interrelationships btwn various exposures and dz
|
|
What is meant by interaction
|
"When the incidence rate of disease in the presence of two or more risk factors differs from the incidence rate expected to result from their individual effects."
|
|
List 2 types of bias:
|
1 Selection bias (exclusion bias)
2 information bias (x6) |
|
What are two things that should be taken into account in regards to confounding when designing a study:
|
1. Ensure that adequate date are collected on possible confounders.
2. impossible to control for confounder when there is not enough data about them |
|
How do you collect data on possible confounders?
|
-impossible to collect info on every varible
-choose based on: knowledge of dz,previous evaluations of the same or related questions,best judgement of investigator at the time the study was initiated -collect data on known risk factors |
|
How to determine if a factor is confounder?
|
analyze available data
- obtain a crude overall estimate of the assoc.,control for the factor,observe whether the estimate of the assoc. btwn the expos. and the dz is changed. -always keep track of which is the exposure and which is the dz and which factor you are evaluating as a possible confounder. |
|
What are some methods for controls for confounding in the design of a study?
|
1) randomization (intervention studies)
2) restriction 3)matching (must maintain in the analysis) |
|
when controlling for confounding using restriction list 5 important points:
|
1) confounding cannot occur if the potential confounding factor do not vary across either the exposure or the dz
2)if concerned about confounding by sex and race limit study to ex: black women,asian men etc) 3)eliminate any possible effects of confounding 4)straight forward and inexpensive 5)limitations |
|
multivariable analysis:
|
1.allows for efficient estimation of measures of assoc. while controlling for a number of confounders simultaneously
2.involves the construct. of a mathmatical model 3.regression and cox proportional hazards model |
|
what is the general regression model?
|
y=mx+b
x=ind. variable y=the dependent variable m describes the relationship between x and Y |
|
Which of the following is an approach to handling confounding?
|
a. individual matching
b. stratification c. group matching d. adjustment e. all of the above |
|
It has been suggested that physicians may examine women who wse OC more often or more thoroughly that women who do not, if so, and if a assoc. is observed between phlebitis and oral contraceptive use the association may be due to:
|
a. selection bias
b. interviewer bias c. surveillance bias d. nonresponse bias e. recall bias |
|
examples of interaction:
|
RR of ca smokers to nonsmoker=3.0
RR of ca abestos exp to nonsmoker=5.0 RR of ca in smoker with asb exp=50.0 -effect is differ. from an addictive model -effect is different from a multiplicative model -suggests something goes on at a biological level |
|
Precision
|
-the degree to which there is variation in a measurement
-reliability refers to the precision of screening and diagnostic tests |
|
accuracy
|
-Refers to the degree to which the measurement is , on average, correct.
-Validity refers to the accuracy of screening and diagnostic tests ex:If each time a phenomenon is measured, the value is the same, but all measurements are far away from the true value, there is high precision but low accuracy This was the case with the class prediction of average age The 95% CI was small, but the true value of the average age was not included -- inaccurate |
|
If chance, bias, and confounding are unlikely explanations of the findings, have est. assoc. BUT NOT.....
|
causality
|
|
Temporal Relationship
|
exposure must precede an outcome for the relationship to be causal
-Not always easily demonstrable b/c sometimes symptoms of dz lead to changes in behavior People who start having resp. problems may stop smoking W/in the year that they stop smoking, they may be diag. with lung ca. May appear as if stopping smoking brings about lung ca. |
|
“confounding by indication”
|
An association based on behavior taken because of the outcome
ex: of temporal as well as confounding; People with kidney insufficiency often progress to total kidney failure People with kidney insufficiency to use acetaminophen (Tylenol) for pain and to avoid aspirin and ibuprofen May look like acetaminophen causes total kidney failure |
|
Strength of Association(p213)
|
-The strength of the association is measured by the relative risk or the odds ratio
-The stronger the assoc., the more likely it is that the relationship is causal -Makes sense – to account for a great increase in risk, there would have to be something really important among the exposed to create such a large increase in risk -Does not imply that an association of small magnitude cannot be judged to be one of cause and effect – only that it is more difficult to exclude alternative explanations |
|
Dose-Response Relationship(p213)
|
-Dose-response rel. that as the dose of the exposure increases, the risk of disease also increases
-Strong evidence of a causal relationship -Absence of dose-response does not rule out a causal relationship -Certain exp. may have a threshold: no dz may develop until a certain threshold is reached |
|
Replication of Findings(p213)
|
If the relationship is causal, we would expect to find it consistently in different studies and in different populations
|
|
Biologic Plausibilityp213
|
-Causality is enhanced if there is a known or postulated biologic mechanism by which the exp. might reasonably alter the risk of developing dz
-What is considered biologically plausible at any time depends on the current state of knowledge Lack of a known or postulated mech.doesn't necessarily mean that a relationship is not causal E.g. scurvy among sailors in the British Navy |
|
Cessation of Exposure p213
|
If a factor is a cause of a disease, expect the risk of dz to decline when the exp. is reduced or eliminated
Eg. Recall the spinach, decrease the rate of E. coli 0157:H7 In some cases, the dz-causing process has already been initiated and dz occurrence may have been det.by the time the exp.is removed |
|
Specificity of the Association p215
|
An assoc. is specific when a certain exp.is assoc.w/ only 1 dz
Vinyl chloride and bladder ca -This is a weak guideline -Absence of specificity in no way negates a causal relationship |
|
Incidence attributable to exposure
|
EXPOSED GROUP
incidence in the exp. group - incidence in the non-exp. TOTAL POP (Incidence in total population) - (Incidence in nonexposed group) |
|
Proportion of incidence attributable to exposure
|
(Iexp - I nonexp)
--------------------- I exp |