Epidemiology/BioStats Flashcards
Prevalence?
of existing cases in population at a specific moment in time.
(Total number of cases in population at one point in time)/(total population)
Incidence?
of new cases in the disease free population that develop over a period of time
(of new cases in the disease free population over a given time period)/(total population at risk during specified time period)
(remember to subtract pre-existing cases of the disease from total population at risk)
Prevalence depends on?
Incidence and duration of disease
P = I X D
Incidence can be measured in a ___ study?
cohort
Prevelance can be measured in a ____ study?
cross-sectional
As the mortality of a disease decreases, prevalence of the disease ______ because ____ increases?
Increases because duration of disease increases.
Sensitivity
Probability that a patient with a disease will have a + test result
sNOUT- will rarely miss people with the disease so it is good at ruling out those who do not have the disease.
Desirable early in diagnostic work up
Specificity
Probability that a patient without the disease will have a negative test result. A specific test will rarely say someone has the disease when they do not so its good at ruling IN.
sPIN - desirable to confirm a diagnosis
The higher the prevalence of a disease, the higher the ____
positive predictive value
positive predictive value
probability that a patient with a (+) test result truly has the disease
negative predictive value
probability that patient with (-) test result truly does not have disease
the lower the prevalence, the higher the?
NPV
Likelihood ratio?
express extent to which a given test result is likely in diseased people vs people without disease
(+) LR shows?
how much the odds of disease are increased if test result is (+)
(-) LR shows?
how much the odds of disease decrease if test result is (-)
Posttest odds?
Pretest odds X LR
Absolute risk?
Incidence of disease
Attributable risk (risk difference)
Difference in risk between exposed/unexposed groups
Incidence of disease in exposed - incidence in unexposed
needed to treat
of individuals that need to be treated for 1 patient to benefit
1/attributable risk
1/(incidence of disease in exposed - incidence in unexposed)
relative risk or risk ratio?
how much more likely an exposed person is to get the disease in comparison to unexposed
(measures strength of association between exposure and disease)
relative risk/risk ratio equation?
incidence in exposed/incidence in unexposed
RR > 1 suggests? RR<1 suggests?
>1 = increased risk <1 = decreased risk
Odds ratio definition and equation
estimates relative risk in case control studies. the lower the disease incidence, the more closely it approximates RR.
odds that diseased person is exposed / odds that non-diseased person is exposed
Kaplan-Meier curve
describes survival in cohort of patients over time
gold standard study to evaluate diagnosis, treatment, etc?
double-blinded randomized control trial
List types of studies in descending order of quality?
RCT, observational study-(used if exposure is therapeutic intervention), case series/case reports.
double - blinded studies prevent _____ bias?
observation
cross-sectional study?
observational study that assesses risk factors and outcomes at snap-shot in time. (measures correlation, not causation)
Prevalence study?
cross sectional study performed to measure prevalence.
cohort study (also known as longitudinal or incidence study)
group of people without outcome of interest (disease) are assembled. For each possible risk factor people are categorized as exposed or unexposed and then they are all followed over time. Incidence of outcome is compared in exposure groups.
(Ex: Framingham heart study)
What study is only way to directly determine incidence of disease?
cohort
2 types of cohort studies?
prospective or retrospective
case-control study
Two groups are identified (case group) and (control group) and then the frequency of a risk factor is compared between them.
case group- heart disease
control- no heart disease
risk- red meat exposure
advantage of case control study?
can be used to study rare diseases as they can use smaller groups than cohort studies
RCT
patients randomly assigned to treatment vs control groups.
ability to demonstrate causal relationships.
Cohort studies can measure?
RR, OR, Incidence, Prevalence
Cross-sectional studies measure?
Prevalence (not incidence) because measurements occur at a specific moment in time.
case control studies measure?
Odds Ratio (approximating relative risk)
What type of bias is particularly problematic in retrospective cohort studies?
selection bias
What types of bias are common in case control studies?
recall bias (since groups with or without disease are selected first and then asked about exposures)
Selection bias
Selection bias
samples/participants selected that may differ from rest of population
ex: individuals concerned about breast cancer due to family history may be more likely to enroll in a study
measurement bias
measure or data gathering diff btwn groups (one group measured with CT, other with MRI)
Confounding bias
3rd variable positively or negatively associated with exposure or outcome (leads to incorrect association)
effect modification
third variable disproportionately affects 2 groups. It shows a meaningful difference (vs confounding which does not)
Ex: new chemo drug shows improved survival for patients undergoing radiation but no effect for patients not undergoing radiation.
lead-time bias
results from earlier detection of disease- gives appearance of prolonged survival
length bias
screening tests detect disproportionate # of slowly progressive disease but miss rapidly progressive ones (leads to overestimation of benefit of the screen)
Type I error:
concluding that there is a difference in treatment when there is not. (false +)
-rejecting null hypothesis when it should not be
Type II error:
concluding no difference in treatment effects when there is a difference. (false negative)
Failing to reject the null hypothesis.
p value
estimate of probability that differences in treatment effects in a study could have happened by chance alone
does p value say anything about direction or size of the effect?
No
Power
probability that a study will find a statistically significant difference when one is truly there
Increasing ____ will increase power of study?
of study subjects
Power equation?
1 - type II error.
Confidence interval
Demonstrates the statistical significance- showing the size of the effect and the statistical power (the narrower the confidence interval- the greater the statistical power)
If the confidence interval contains the (RR or OR of 1.0 or 0%), the results are?
not statistically significant
Live-attenuated vaccines?
Measles, mumps, rubella, polio, yellow fever, influenza (nasal spray)
Toxoid vaccines
Diptheria, tetanus
Subunit vaccines
HBV, pertussis, strep pneumo, HPV, meningococcus
Inactivated (killed) vaccines
cholera, HAV, polio (salk), rabies, influenza (injection)
conjugate
Hib, Spneumoniae
Name the stages of change in behavioral counseling from beginning to end
Precontemplation Contemplation Preparation Action Maintenance
Precontemplation
denial or ignorance of problem
contemplation
ambivalence or conflicted emotions, assessing benefits and barriers
preparation
experimenting with small changes. collecting info about change
action
taking direct action toward a goal
maintenance
maintaining a new behavior, avoiding temptation
Name stage of contemplation.
A heroin addict visits doctor to ask questions about quitting.
preparation
addict enters rehab facility for treatment of addiction.
action
a heroin addict has never thought about cessation.
precontemplation
heroin addict considered treatment for addiction
contemplation
heroin addict continues to visit recovery meetings to gain support and reinforcement against relapse
maintenance
What age can you stop mammogram screening?
75
When can you stop colon cancer screening?
75
What is difference in cholesterol screening btwn men and women?
men: screen at age 20 for patients with increased risk of heart disease. screen all men above age 35.
w: screen for women at increased risk
When should AAA screening be performed?
One time screening in all men over 65 who have ever smoked.
leading cause of cancer mortality in US?
lung cancer
most prevalent cancers in women? men?
breast, prostate
lung, colorectal rank 2nd and 3rd in both
If a step 2 question asks you to calculate relative risk from retrospective data, the answer is?
cannot be calculated!
none of the above!
(relative risk can only be calculated from prospective or experimental studies)
What is a clinically significant value for relative risk?
Anything other than 1.
1.5 = 1.5X more likely to develop condition if exposed to the factor in question
.5 = only 1/2 as likely to develop condition (factor is protective)
With a normal bell shaped distribution, 1 SD holds ___ % of values, 2 SD holds? 3 SD holds?
1 SD = 68% of values
2 SD = 95%
3 SD = 99.7%
If the mean score on a test is 80 and standard deviation is 5, what percent of scores will be within 5 points?
what percent of scores will be within 10 points?
15 points?
68%
95%
99.7%
when do mean = median = mode?
in a normal distribution
positive skew distribution?
how does this affect mean, median, mode relationship?
asymmetric distribution with an excess of high values (tail of curve to the right)
mean > median > mode
negative skew?
excess of low values (tail to left of curve)
mean < median < mode
test reliability?
synonymous with test precision. measures reliability and reproducibility of test. (person taking test will get same score with 2 diff people administering it)
validity of test?
accuracy of test. does test measure what it claims to. when a genius takes IQ test, score should reflect genius status.
correlation coefficients?
range from -1 to +1. measure the degree to which 2 variables are related.
is a correlation of -.6 or .4 stronger?
-.6 (stronger negative correlation)
Define confidence interval?
When you take data from subset of population and calculate its mean, you want to say that its equivalent to the mean of the whole population. A confidence interval of 95% suggests that you are 95% confident that the mean of the whole population falls within 2 SD of your sample mean.
retrospective studies are more practical for ____ disease?
rare
prospective studies are more practical for ____ diseases?
common
If a disease can only be treated to the point that people can be kept alive longer without being cured, what happens to incidence and prevalence?
Incidence- stays the same. same # of people contract the disease every year.
Prevalence- increase because ppl live longer
In short term diseases (flu), incidence will be higher or lower than prevalence?
higher
in long term diseases (diabetes), incidence will be higher or lower?
lower
epidemic
observed incidence greatly exceeds expected incidence
When do you use chi-squared test?
compare percentages or proportions (non-numeric/nominal or ordinal data)
t-test?
compares 2 means
analysis of variance (ANOVA)
compares 3 or more means
What is the difference between nominal, ordinal and continuous types of data?
nominal- no numeric value (day of week)
ordinal- give ranking but no quantity (clas rank- does not specify how far ahead 1 is from 2)
continuous- most numerical measurements (weight, BP, age)
Chi-square test must be used to compare what kind of data?
nominal - non numeric
ordinal- ranking (no quantity)
t-test and ANOVA are used to compare?
continous data
What does it mean if P is less than .05?
there is less than a 5% change that data were obtained by random error or chance
If P is less than .01?
there is less than a 1% chance that data were obtained by random error/chance
3 important points to remember about P values for Step 2:
- A study with a P-value less than .05 can still have serious flaws
- a low P value does not imply causation
- A study that has statistical significance does not necessarily have clinical significance.
What is the relationship of the p-value to the null hypothesis?
When P is less than .05 I can confidently reject the null hypothesis. There is less than a 5% chance that the null hypothesis is correct.
The p-value represents the chance of making what type of error?
Type I error (saying there is an effect when in fact there is not)
best way to increase power?
increase sample size
unacceptability bias?
people do not admit to embarrassing behavior. claim to exercise more than they do. claim to take experimental meds then they are actually spitting them out.
Interventions targeting which group have been shown to have a significant effect in improving quality of patient care when it comes to managing medications?
Interventions targeting pharmacists and high risk patients
Receiver operating characteristic curve
plot true positive rate (sensitivity) against 1- true negative rate (specificity)
Area under a receiver operating characteristic curve?
= Accuracy
Accuracy of a test equals?
proportion of true results (true positive and true negative out of all the results. depends on specificity, sensitivity, prevalence of disease
What is the rare disease assumption?
If a disease is rare- low prevalence and low incidence, then the Odds ratio can generally approximate the RR.
When can the odds ratio more reliably approximate relative risk?
- When incidence of disease is low
2. When cases and controls used in a study are representative of indv with and without disease in gen population
Hawthorne effect
tendency of a study population to affect outcome because the people are aware that they are being studied.
sample distortion bias
estimate of exposure and outcome is biased because study sample is not representative of target population
information bias
due to imperfect assessment of association between exposure and outcome measurements
Typically, lowering a cutoff value will ______ sensitivity?
Increase
When a disease has few false negatives, this means it has a high
sensitivity
when a disease has few false positives, this means it has a high
specificity.
if the PPV of a test is 25% and a patient is found to have a positive result, is their post-test probability of having the disease high or low?
low
False positive ratio =
1 - specificity
False negative ratio =
1 - sensitivity
When the cut off level of a diagnostic test is raised, the ____ rate with consequently increase
false negative rate
observer bias
occurs when investigators decision is adversely affected by knowledge of the exposure status
respondent bias
outcomes are obtained by patient response and not by objective diagnostic measurements (blood pressure)
Factorial design study exhibit
involves 2 or more experimental interventions. each with 2 or more variables that are studied independently.
cluster analysis
grouping of different data point into similar categories
cross-over study
group participants assigned to 1 treatment for a period of time and then switch treatments for another set period of time.
parallel study
randomizes one treatment to 1 group and a different treatment to the other group
effect modification
external variable positively or negatively impacts the effect of risk factor on disease (example venous thrombosis increases with estrogen therapy and this is further augmented by smoking)
post-hoc analysis
refers to performing unplanned statistical tests on patterns that were identified after the fact
selective survival bias
occurs in case control studies when cases are selected from entire disease population rather than newly diagnosed cases.
A relative risk of .71 for a new drug measuring outcomes on heart failure says?
The new drug decreased heart failure risk by .29 or 29%
measurement bias occurs from?
poor data collection with innaccurate results
Form of selection bias known as susceptibility bias
when treatment regimen selected for a patient depends on severity of their condition and not on the group they were assigned to
Cholesterol levels are measured for a test 3 times and come back as 200, 190, 184. This means the test is not?
reliable
How do you determine validity of a test?
tests ability to measure what its actually supposed to measure. So you compare results to the gold standard.
generalizability or external validity
pertains to applicability of study results to other populations
