Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

Epidemiology > Evaluating the Role of Random Error > Flashcards

Evaluating the Role of Random Error Flashcards

1
Q

If the observed result is false, what could be the 3 alternate explanations?

A
  • Bias
  • Confounding
    • Random Error
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Random Error

A
  • aka chance
  • arises from:
    • measurement errors
    • sampling variability
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are some ways to reduce random error?

A
  • increase sample size
  • repeat a measurement or entire study
  • use an efficient study design
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Hypothesis Testing

A
  • determines if random error/chance causes association
  • P value
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Null Hypothesis

A
  • no association b/w exposure and outcome
  • RR=1, OR=1, RD=0
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Alternate Hypothesis

A
  • is an association b/w exposure and outcome
  • RR≠1, OR≠1, RD≠0
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Define P Value:

A
  • given that H0 is true, the p-value=probability of observed and extreme results by chance only
  • Range: 0-1
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

If p ≤ .05…

A
  • significant
  • reject H0, Accept HA
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

If p>0.05

A
  • not significant
  • do no reject H0
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Problems with use of P-value in research

A
  • Does not:
    • imply medical, biological, or public health significance
    • rule out bias or confounding explanations
    • mean H0 is true
  • P≤0.05 is arbitrary→ judgement errors
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What does the confidence interval tell you

A
  • range of hypotheses that are comparable to the data
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What does a wide confidence interval indicate?

A
  • smaller population size
    • large amount of random error
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

what does a narrow confidence interval indicate?

A
  • large sample size
    • small amount of random error
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

If null hypothesis is within confidence interval range, the what would the p-value be?

A
  • P-value> 0.05
  • not significant
    • H0 not rejected
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Confounding

A
  • mixing of effect b/w exposure, outcome, and confounder (3rd variable)
  • exaggerates or minimizes true association
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Criteria to be a confounder?

A
  • Must be associated with:
    • exposure
    • outcome
      • independent of exposure
  • not intermediate step in cause pathway b/w exposure and disease
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What are the effects of confounding?

A
  • account for all or part of association
  • cause overestimate or underestimate of association
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What variables can be potential confounders?

A
  • risk factors for disease
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What are the different ways to control for confounding?

A
  • Design Stage:
    • Randomization (RCT)
    • Restriction
    • Matching
  • Analysis Stage:
    • Stratification
    • Multivariat analysis
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Randomization

A
  • with sufficient sample size
    • control for known and unknown confounders
  • not guaranteed
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

Restriction

A
  • restrict admission criteria for study
  • limits individuals in specific category of confounder
  • ex: Race limited by age or height
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What is the goal for controlling confounding in design phase

A
  • eliminate or reduce variation in confounding factor b/w compared groups
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

Restriction: Advantages and disadvantages

A
  • advantages
    • straight forward
    • convenient
    • inexpensive
  • disadvantages
    • limits generalization
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

Matching

A
  • select subject so potential confounders are distributed in identical manner among:
    • exposed and unexposed
      • cohort study
    • case and controls
      • case control study
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

Stratification

A
  • evaluate association within homogenous categories (strata) of confounding variable
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
26
Q

Bias

A
  • systematic error
  • results in and incorrect measure of association:
    • creates association when there is not one
      • bias away from null
    • Mask association when there Is one
      • bias towards the null
  • arise in all study types:
    • experimental
    • cohort
      • retrospective=mostsusceptible
    • case-control
  • evaluated but not fixed in analysis design
27
Q

What are the 2 main type of bias?

A
  • selection
  • observation
28
Q

What type of studies are Selection Bias most likely to occur in? Why?

A
  • most likely to occur in:
    • case control
    • retrospective
  • bc Exposure and outcome occurred at time of study selection
29
Q

What is the solution to selection bias?

A
  • nothing can be done
30
Q

Observation Bias

A
  • error in the info on exposure or outcome from subjects
  • affects groups unequally
  • occurs after subject have entered the study
  • results in incorrect classification of subjects:
    • exposed vs unexposed
    • diseased or not
31
Q

What are the different types of observation bias?

A
  • Recall
  • Interviewer
  • misclassification
32
Q

recall bias

A
  • remember or report exposures differently
    • more or less accurate than those w/o disease
33
Q

Interviewer bias

A
  • difference in soliciting, recording or interpreting info
34
Q

misclassification bias

A
  • subjects exposure or disease status is classified wrong
    • incorrectly reported or recorded
35
Q

Loss of subjects to f/u

A
  • Problem in cohort studies when related to exposure and disease
36
Q

What diseases are appropriate for screening?

A
  • serious disease w/severe consequences
  • progressive disease & tx is more effective at earlier stage
  • High Prevalence in DPCP
    • Detectable pre-clinical phase
37
Q

What are some characteristics of good screening test?

A
  • economical
  • convenient
  • free of risk and discomfort
  • acceptable to large population
  • highly valid and reliable
38
Q

test validity

A
  • does it accurately identify those with and without preclinical disease
    • sensitivity
    • specificity
39
Q

Test reliability

A
  • same result each time
40
Q

2x2 table for screening test outcomes:

A
41
Q

Define Sensitivity

A
  • probability that people with the disease will test positive
    • True Positive
42
Q

How do you calculate sensitivity?

A

a/(a+c)

43
Q

Define Specificity

A
  • probability that people without the disease have a negative test
    • true negative
44
Q

t

A
45
Q

How do you calculate specificity?

A

d/(b+d)

46
Q

define Positive Predictive value

A
  • the likelihood that a positive test result indicates the existence of the disease
    • how worried should the patient be?
47
Q

How do you calculate positive predictive value?

A

a/(a+b)

48
Q

define negative predictive value

A
  • likelihood that a negative test result indicates the absence of the disease
    • how reassured should they be?
49
Q

How do you calculate negative predictive value?

A

d/(c+d)

50
Q

What do predictive values depend on?

A
  • sensitivity
  • specificity
  • prevalence
51
Q

Randomized control Trials

A
  • gold standard for study designs
    • bc of greater internal validity
52
Q

Steps to conduct RCT

A
  • enroll subject
  • allocate to tx groups (Tx vs placebo)
    • randomization
    • blinding
  • F/u for relevant period
  • study groups monitored for outcome
    • compare rates of outcome in different groups
53
Q

RCT Step 2: Allocate Tx or placebo by randomization. What are the benefits? Overall Goal?

A
  • Benefits:
    • unbiased assignment
    • known and unknown confounders are balance
    • leads to comparability
    • minimizes selection bias and confounding
  • Goal:
    • Groups are identical-all determinants except tx
      • any difference due to tx
54
Q

RCT: What groups are subjects randomized into?

A
  • Tx Group
    • New tx
    • other doses of same tx
    • Other: Combo therapy
  • Comparison group
    • Placebo
    • Other: usual care (current tx in use)
55
Q

What is the primary objective for using a placebo?

A
  • minimize bias
56
Q

Uses of placebo?

A
  • makes group as comparable as possible
  • depends on subjectivity of outcome
  • cannot be done in some situations
    • pills vs surgery
  • Allows study to be blinded
57
Q

What is the primary objective of blinding in RCT?

A
  • Prevent bias from affecting results
58
Q

What are the different types of blinding?

A
  • Single-Blind
    • subjects unaware of treatment group
  • Double-Blind
    • Both subjects & investigators unaware of tx groups
      • cannot be done in some situations (pills vs surgery)
  • Triple Blind:
    • Subjects, investigators admin tx, investigators eval outcomes all unaware
59
Q

How to establish cause

A
  • Valid Association
    • even if valid, not all associations are causal
  • eliminated alternative explanations
60
Q

Hill’s Criteria

A
  • Causality is strengthened by:
    • temporal sequence
      • risk factor first
    • strong association
    • dose-response relationship
    • consistent findings
    • biologic credibility
    • results of experiment
61
Q

Temporal sequence

A
  • a factor must come before outcome to cause it
62
Q

What types of studies provide the best evidence that causes precedes the effect?

A

Prospective Cohort Studies

Randomized Clinical trials

63
Q

Strength of Association

A
  • Large RR supports cause relationship
    • by itself, can’t conclude cause
      • bias or confounding
  • Small RR also possible
64
Q

Presence of Dose-Response relationship

A

more exposure means greater risk

Epidemiology (6 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions