Epidemiology And Biostatistics - Slumd

Front 1

Epidemiology

Back 1

The study of how disease is distributed in populations and the factors that influence or determine this distribution

Front 2

5 Goals of Epidemiology

Back 2

IDSEP
Identify causes and risk factors
Determine extent of disease
Study the natural history of disease
Evaluate preventative and theraputic interventions
Provide foundation for public policy interventions.

Front 3

Incidence

Back 3

Number of new cases occurring in a particular time period.

Front 4

Incidence rate

Back 4

the ratio of new cases ocurring in a particular time period to the total number of people AT RISK

Front 5

Point Prevalence

Back 5

number of people affected by a disease at a particular point in time.

Front 6

Period prevalence

Back 6

The total number of people affected by a disease over a particular time period.

Front 7

Prevalence rate

Back 7

ratio of the number of people with a disease to the number of people at risk at a particular time or time period.

Front 8

Attack Rate

Back 8

ratio of the number of people contracting a disease to the number of people at risk (expressed as a percentage)

Front 9

What is the difference between incidence rate and attack rate.

Back 9

Attack rate is used to describe a short term risk or exposure (Like the seasonal flu)

Front 10

Mortality rate

Back 10

ratio of the number of people dying over a period of time to the number of people at risk

Front 11

What is the effect of preventative measures on Prevalence and Incidence?

Back 11

decrease both incidence and prevalence

Front 12

Case Fatality Rate

Back 12

the ratio of the number of people dying in a particular episode of a disease to the total number of episodes of the disease expressed as a percentage.

Think Pneumonia. 20% of pneumonia cases are fatal for example.

Front 13

Treatment of a disease with full recovery will do what to incidence and prevalence?

Back 13

Prevalence decreases, incidence remains constant.

Front 14

Theraputic treatment of a disease does what to incidence and prevalence?

Back 14

Increases prevalence, incidence remains constant.

Front 15

Why is prevalence not always the best statistic to analyze the success of a treatment?

Back 15

Prevalence will increase in cases of death prevention or theraputic treatment.

Front 16

Name the two broad categories of data.

Back 16

Discrete - groups or ranks
Continuous - intervals or values

Front 17

Nominal Data

Back 17

Discret groups: male/female

Front 18

Ordinal Data

Back 18

Discrete data ordered without meaningful intervals.

The good
The bad
The ugly

Front 19

Interval Data

Back 19

Continuous data with meaningful intervals not centered around an absolute 0. (Temperature in celsius)

Front 20

Ratio Data

Back 20

Interval data with an absolute 0

Front 21

Frequency

Back 21

absolute number in a particular category

Front 22

Relative frequency

Back 22

percentage of values in a particular category when compared to the whole

Front 23

Cumulative frequency

Back 23

Total number at or below a threshold. Percentile.

Front 24

Frequency Polygon

Back 24

used to display continuous interval data where the midpoint of each subgroup is marked as a point and connected by lines.

Front 25

Cumulative Frequency polygon

Back 25

Similar to frequency polygon, but the values are additive and eventually reach a maximum point in the last subgroup.

Great for estimating percentile data.

Front 26

What is the benefit of a Kaplan Meier Curve when compared with a normal survival curve?

Back 26

Kaplan Mayer curves are referred to as censored data. This accounts for new additions or early exits from the study. (For example, a car accident in a heart surgery survival study)

cuts from both numerator and denominator.

Front 27

Positive Skew

Back 27

Varient of normal distribution with a long tail to the right signifying a number of extreme values which are higher than the mean.

Mode --> Median --> Mean

Front 28

Right Skew

Back 28

Varient of normal distribution with a long tail to the right signifying a number of extreme values which are higher than the mean.

Mode --> Median --> Mean

Front 29

Left Skew

Back 29

Varient of normal distribution with a long tail to the left signifying a number of extreme values which are lower than the mean.

Mean-->Median-->Mode

Front 30

Negative Skew

Back 30

Varient of normal distribution with a long tail to the left signifying a number of extreme values which are lower than the mean.

Mean-->Median-->Mode

Front 31

Variance

Back 31

quantifies the scatter (spread) present in a distribution of values.

average of the squared differences from the mean.

Front 32

Standard deviation

Back 32

square root of the variance

68% of data is within 1 SD
95% within 2 SD
99.7% within 3 SD

Front 33

What percentage of data is within 1 SD?

Back 33

68%

Front 34

What percentage of data is within 2SD?

Back 34

95%

Front 35

What percentage of data is within 3 SD?

Back 35

99.7%

Front 36

What percentage of data is above 2SD in a normal distribution?

Back 36

2.5% ABOVE!!!

Front 37

What is the probability of two independent events occuring?

Back 37

P1 X P2

Front 38

What is the probability that 1 of two possible outcomes occur?

Back 38

P1 + P2 - (P1xP2)

Remember to exclude the possibility of both happening.

Front 39

Parametric Tests

Back 39

focused on population parameters.
require a continuous variable

(assumes a normal distribution)
Think p values!

Front 40

Sampling distribution.

Back 40

Averages data in a bimodal distribution to create a normal distribution.

Generated by the central limit theorum.

Front 41

non parametric tests

Back 41

can be applied to discrete variables. does not make assumptions about the distribution.

Front 42

Paired t-test

Back 42

Parametric - compares mean of a single group before and after treatment

Front 43

T-test

Back 43

Compares means between two groups. (experimental vs placebo)

Front 44

ANOVA

Back 44

parametric test - compares means between more than two groups.

Front 45

Chi-squared

Back 45

non parametric test - a test of proportions.

Front 46

Type I error

Back 46

False Positive

Front 47

Type II error

Back 47

False Negative

Front 48

Power

Back 48

estimate of the ability of a study to detect a false null hypothesis. in other words, to detect a significant difference.

the probability of avoiding a type II error.

Front 49

what is the Power of a test with an beta value of .05 that rejects the null hypothesis

Back 49

95%

Front 50

Standard error of the mean.

Back 50

found by dividing the stardard deviation by the square root of the sample size.

Statistically misleading. used to fudge data.

Front 51

Randomized Controlled Trials

Back 51

subjects randomly assigned to interventions groups with a control group involved.

Front 52

List the three types of Randomization discussed in class and describe them.

Back 52

Pure random selection - equal probabiltiy of anyone being chosen

Systematic Randomization - every third infant born...

Stratified randomization - divide groups to ensure equal representtion of sub groups.

Front 53

Intention to Treat Analysis

Back 53

Each subject is assumed to comply with the study regardless of cooperation.

If a drug has horrible side effects and the subject withdraws, this will be reported as if the patient took the drug. (enables failure to be acocunted for)

Typically this study is done in conjunction with a study and not alone.

Front 54

Interim analysis

Back 54

Analyzing data before the scheduled end of the study. done for ethical concerns of the subjects.

Front 55

Prospective Cohort Studies

Back 55

observing a cohort not affected by any disease and recording data on their risk factors and future exposure.

Front 56

What is a major flaw of prospective cohort studies?

Back 56

It is not effective for rare diseases and expensive.

Front 57

Inception Cohort

Back 57

subjects followed soon after developing disease.

Front 58

Historical Cohort

Back 58

information regarding the cohort is obtained from previously collected or existing historical data.

Front 59

Case Control Study

Back 59

a group affected by a condition is compared with a group unaffected. Retrospective data on risk factors is then collected

Front 60

Cross Sectional Study

Back 60

survey a population for simultaneous presence of a disease and potential risk factors.

Prevalence of diabetes in one community compared with another.

Front 61

Can causality be determined in a cross sectional study?

Back 61

No, only suspected since symptoms and risk factors are discovered simultaneously.

Front 62

Case Report/ Case Series

Back 62

describes a clinical event in a single patient or series of patients.

no statistical validation is possible.

Front 63

Systematic Review

Back 63

summary of the medical literature regarding a clinical question.

Most powerful study if quality is high.

Front 64

What are the potential pitfalls of a systemic review?

Back 64

Heterogeneity and Publication Bias.

Front 65

Gold Standard

Back 65

a test which is considered to be consistently correct and to which other tests can be compared.

Front 66

Reliability

Back 66

level of agreement between repeated measurements of the same variable- reproducability

Front 67

Validity

Back 67

extent to which a test actually tests for what it claims to test.

Front 68

Sensitivity

Back 68

ability to detect people who have the disease.

TP/(TP+FN)

Front 69

Specificity

Back 69

Ability to detect people who do not have the disease

Front 70

what does SPIN refer to?

Back 70

Tests with high specificity can be used to rule in outcomes

Front 71

What does SNOUT refer to?

Back 71

Tests with high sensitivity can be used to rule out outcomes

Front 72

How do you calculate sensitivity?

Back 72

Sensitivity = True positives / (True positives + False Negatives)

Front 73

How do you calculate specificity

Back 73

Specificity = True Negatives/ True negatives + False Positives.

Front 74

Positive Predictive value

Back 74

liklihood that a positive test actually corresponds to the disease

PPV = TP/(TP+FP)

Front 75

Negative predictive value

Back 75

likelihood that a person with a negative result does not have the disease.

NPV = TN/(TN+FN)

Front 76

How do you calculate Prevalence?

Back 76

True positives + False Negatives / All outcomes

Front 77

How do you calculate positive predictive value?

Back 77

True positives / True positives + False Postives

Front 78

how do you calculate negative predictive value?

Back 78

True negatives / True negatives + False Negatives

Front 79

What factor needs to be considered in order to use Predictive Values?

Back 79

Prevalence. If prevalence is not know the predictive values will be snapshot estimates rather than statistical information of the general population.

Front 80

What is the likelihood ratio.

Back 80

A statistical representation of the selectivity and sensitivity of a test that can be used to compare pre test and post test outcomes via a simple line.

Front 81

ROC curve

Back 81

plot the true positive rate (sensitivity) against the false positive rate (1- specificity) for different cutoffs of a diagnostic test

Front 82

Likelihood ratio

Back 82

liklihood that a given test result would be expected in a patient with the target disorder compared to the liklihood that that same result would be expected in a patient without the target disorder.

Front 83

An r value of .55 corresponds to what strength of correlation?

Back 83

strong.

Front 84

Which r value demonstates a stronger correlation .55 or -.60?

Back 84

-.60 This is a stronger negative correlation

Front 85

how does one obtain the coefficient of determination? What does it signify?

Back 85

r squared - it is the amount of variance explained by a particular variable.

Front 86

Odds ratio

Back 86

measures the degree of association of a risk factor with a disease or outcome. It is the ratio of the odds that a case was exposed to the odds that a control was exposed.

Used when actual incidence is not measured

Front 87

Absolute Risk

Back 87

synonymous with incidence rate. Number of people developing the disease / number of total people.

Front 88

Relative Risk

Back 88

measures how many times exposure to a risk factor increases the risk of contracting a disease. it is the ratio of the absolute risk of disease among those exposed to the absolute risk among those not exposed.

Can only be measured when incidence is measured.

Front 89

What is the formula for RR?

Back 89

Relative Risk = AR exposed / AR control

Front 90

What is RRR?

Back 90

Relative risk reduction = 1-RR(post treament)

Front 91

Absolute Risk Reduction

Back 91

decrease in absolute risk due to treatment expressed as a percentage.

Front 92

Why is ARR more clinically relevant than RRR?

Back 92

if the risk of contracting an extremely rare disease is reduced 50% (RRR) it will look more impressive than its ARR which would be 1/200,000,000

Front 93

how do you calculate NNT

Back 93

1/ARR

Front 94

What is NNT?

Back 94

Number needed to treat, this indicates the number of people needing to receive a treatment in order to produce one positive result.

NNT = 1/Absolute Risk Reduction

Front 95

Cost of Intervention

Back 95

NNT x Cost

Front 96

BICEP!!!

Back 96

Bias
Internal Validity
Confounding Variables
External Validity
Power

Front 97

Bias

Back 97

factors which shift data in a particular direction

Front 98

Internal Validity

Back 98

is the assessment measuring what it intends to?

Front 99

Confounding Variables

Back 99

variables which the research failed to measure that are affecting results

Front 100

External Validity

Back 100

are the studies results valid to the real world

Front 101

Power

Back 101

ability to resist type II error.

is the study able to detect an association if it exists?

Front 102

Compare and Contrast Surrogate Markers and Clinical Outcomes.

Back 102

Surrogate markers are measurable values indicative of a disease whereas clinical outcomes are more significant like MI or death.

Front 103

What are the benefits of using composite outcomes?

Back 103

It increases the chance of finding a statistically significant result in your data.

Increases Power

Front 104

What is a pitfall of composite outcomes

Back 104

It can be misleading if the individual components are not eqaully relevant.

For example... if the study indicates a 60% risk of death when death really is rarely or never happening in the study,

Front 105

How do you grade a study?

Back 105

With a damn table.