Epidemiology and biostatistics Flashcards
Types of observational studies
- Cross-sectional
- Case-control
- Cohort
- Twin concordance
- Adoption
OR is used in
Case-control studies
RR is used in
Cohort studies
-Frequency of disease
-Frequency of risk related factors
Assessed in the present
Cross-sectional study: What is happening now?
Patients with COPD had higher odds of a history of smoking than those without COPD. What observational study did we use?
Case-control study
Smokers had a higher risk of developing COPD than nonsmokers. What observational study did we use?
Cohort study
Compares frequency with wich both monozygotic twins vs both disygotic twins develop the same disease
Twin concordance study
Compares siblings raised by biological vs adoptive parents
Adoption study
Blinding of the researchers analyzing the data
Triple-blind
Phases of a clinical trial
Phase I: is it safe?
Phase II: Does it Work?
Phase III: Improvements?
Phase IV: Market. Can it stay?
Can a cross-sectional study show association between a risk factor and a disease?
It can show association but it does not establish causality
Measures heritability and influence of environmental factors
Twin concordance study
Adoption study
Study sample used in phase I
Small number of healthy volunteers
Study sample used in phase II
Small number of patients with disease of interest
Study sample used in phase III
Large number of patients randomly assigned either to the treatment under investigation or the best available treatment or placebo
ITS THE ONLY ONE USING LARGE NUMBERS
Study sample used in phase IV
Postmarketing surveillance of patients after treatment is approved
Phase I of a clinical trial assesses…
Assesses safety, toxicity, pharmacokinetics and pharmacodynamics
Phase II of a clinical trial assesses…
Treatment efficacy
Optimal dosing
Adverse effects
Phase III of a clinical trial assesses…
Compares the new treatment to the current standard of care
Phase IV of a clinical trial assesses…
Rare or long term adverse effects
Formula of sensitivity
TP/(TP+FN)
If sensitivity test is negative, it…
rules OUT a disease
Proportion of all the people with disease among all those who test positive
Sensitivity
True positive rate
Sensitivity
True negative rate
Specificity
If specificity test is positive, we…
Rule IN disease
If sensitivity is 100%, what is the % of false negatives?
0% of false negatives. So all negatives are True Negatives
If specificity is 100%, what is the % of false positives?
0% of false positives. So all positives are True positives
The probability that when the disease is absent the test is negative
Specificity
The probability that when the disease is present the test is positive
Sensitivity
Test used for screening in diseases with low prevalence
Test with high Sensitivity
Test used for confirmation after a positive screening test
Test with high specificity
Fixed properties of a test
Sensitivity
Specificity
Variables that change depending on disease prevalence in population being tested
Positive and negative predictive values (PPV and NPV)
Define PPV
Proportion of all positive tests that are true positive.
If my results are positive, what are my chances of truly having the disease?
Define NPV
Proportion of all negative tests that are true negative.
If my results are negative, what are my chances of truly not having the disease?
Varies directly with pretest probability
PPV
Varies inversely with prevalence or pretest probability
NPV
Prevalence in low NPV
High
Prevalence in high PPV
High
Lowering the cutoff point of a test
Increases sensibility
Decreases specificity
Decreases PPV
Increases NPV
Rising the cutoff point of a test
Increases specificity
Decreases sensibility
Rises PPV
Decreases NPV
1 - false negative rate
Sensibility
1 - false positive rate
Specificity
LR+
= True positive rate / False positive rate
LR -
= False negative rate / True negative rate
LR+ and LR- indicators of a very useful diagnostic test
LR+ >10
LR- <0.1
RR>1
Exposure associated with more disease occurrence
RR<1
Exposure associated with less disease occurrence
RR=1
No association between exposure and disease
Risk of developing disease in the exposed group divided by the risk of developing disease in the non exposed group
Relative Risk
What happens with RR and OR when disease prevalence is low?
OR approximates RR
Attributable risk, Definition
The difference in risk between exposed and unexposed groups:
the % of disease occurrences that are attributable to the exposure
Relative Risk Reduction (RRR). Formula
RRR= 1 - RR
RRR definition
The proportion of risk reduction attributable to the intervention as compared to a control
Calculate relative risk reduction:
3% of pacients who recieve a flu shot develop the flu, while 10% of unvaccinated patients develop the flue
3/10= 0.3 1-0.3= 0.7= 70% of relative risk reduction
Absolute risk reduction definition
The difference in risk (in absolute terms, not in proportion) attributable to the intervention as compared to a control).
Calculate absolute risk reduction:
3% of pacients who recieve a flu shot develop the flu, while 10% of unvaccinated patients develop the flue
10%-3%= 7%=0.07
Number needed to treat, definition
Number of patients who need to be treated for 1 single patient to benefit
Number needed to treat, formula
NNT=1/ARR
Number needed to harm, definition
Number of patients who need to be exposed to a risk factor for 1 patient to be hARmed
Number needed to hARm, formula
NNH=1/AR
NNT for a good treatment
Low
NNH for a good treatment
High
Incidence rate
Number of new cases / Total people at risk
During a specific period
Prevalence
Number of existing cases / Total population
During a point in time
For a short duration disease, prevalence and incidence…
offer similar results
For chronic diseases, prevalence and incidence
differ, being prevalence larger than incidence
The consistency and reproducibility of a test
Precision
The truness of test measurments
Accuracy
The absence of systematic error
Accuracy
The absence of random error
Precision
Validity
Accuracy
Reliability
Precision
Higher precision
Less standard deviation
More statistical power
Statistical power
1- beta
Selection bias: definition
Non random sampling: study population is not representative of target populated
Berkson bias
Selection bias: study population selected from hospital is less healthy than general population
Healthy worker effect
Selection bias: study population is healthier than the general population
Non response bias
Selection bias: participating subjects differ from nonrespondents in meaningful ways
Strategies to reduce selection bias
Randomization
Ensure the choice of the right comparison/reference group
Bias performing the study
- Recall bias
- Measurment bias
- Procedure bias
- Observer-expectancy bias
Recall bias
Awareness of disorder alters recall by subjects
Recall bias common in
Retrospective studies
Strategies to reduce recall bias
Decrease time from exposure to follow-up
Measurment bias
Information is gathered in a systematically distorted manner
Hawthorne effect
Measurment bias: participants change their behavior in response to their awareness of being observed
Strategies to reduce measurment bias
Use objective, standardized and previously tested methods of data collection that are planned ahead of time
Use placebo group
Procedure bias
Subjets in different groups are not treated the same
Observer-expectancy bias
Researcher’s belief in the efficacy of a treatment changes the outcome of that treatment
Strategies to reduce procedure bias and observer expectancy bias
Blinding
Use of placebo
Bias interprenting results
- Confounding bias
2. Lead-time bias
Confounding bias
When a factor is related to both the exposure and the outcome but not the causal pathway
Lead-time bias
Early detection is confused with increase in survival
Strategies to reduce confounding bias
Multiple studies Crossover studies Matching Restriction Randomization
Crossover studies
Subjects act as their own controls
Strategies to reduce time lead bias
Measure back-end survival: adjust survival according to the severity of disease at the time of diagnosis
Measures of central tendency
Mean
Median
Mode
Mean
Sum of all values / total number of values
Mean affects mostly
Outliers: extreme values
Mode
Most common value
Least affected by outliers
Mode
Median
Middle value of a list of data sorted from least to greatest
Measures of dispersion
Standard deviation
Standard error
Standard deviation
how much variability exists in a set of values, around the mean of these values
An estimate of how much variability exists in a theoretical set of sample means around the true population mean
Standard error
Mean=Median=Mode
Normal distribution, Gaussian
Non normal distributions
Bimodal
Positive skew
Negative skew
Non normal distribution that suggests two different populations
Bimodal
Mean > Median >Mode
Positive skew
Mean
Negative skew
Null hypothesis (H0)
Hypothesis of no difference or relationship
Alternative hypothesis (H1)
Hypothesis of some difference or relationship
Type I error
Stating that there is an effect when none exists
Null hypothesis rejected in favor of alternative hypothesis
Type I error
Alpha
The probability of making a typpe I error
False-positive error
Type I error
Type II error
Stating that there is not an effect when one exists
Null hypothesis not rejected when it is in fact false
Type II error
Beta
The probability of making a type II error
Related to statistical power
Beta: 1- beta = statistical power
Probability of rejecting the null hypothesis when it is false
Statistical power= 1 - beta
Lower beta - Higher statistical power
- Higher precision of the test!
- Higher sample size
- Higher expected effect size
Confidence interval
Range of values within which the true mean of the population is expected to fall, with a specified probability
Standard error formula
Standard deviation / √n
With a higher sample number, SE
Decreases
Z for 95% CI
1.96
Z for 98% CI
2.58
t-test
Checks differences between means of 2 groups
Tea is meant for 2
Comparing mean blood pressure between men and women
t-test
ANOVA
Checks differences between eans of 3 or more groups: Analysis of Variance
Chi-square
Checks differences between 2 or more percentages of categorical outcomes
Positive r value in pearson correlation coefficient
Positive correlation
Negative r value in pearson correlation coefficient
Negative correlation
Coefficient of determination
r square: amount of variance in one variable that can be explained by variance in another variable
