FINAL EXAM Flashcards
(208 cards)
1
Q
The study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to control of health problems
A
The definition of epidemiology according to class lectures
2
Q
True or False
Dr. Garn feels that Gordis’ definition of epidemiology is too narrow and that it should include all aspects of health and well-being and NOT just focus on disease
A
True
3
Q
Infectious disease such as pneumonia and influenza was the leading cause of mortality in 1900. Chronic disease such as heart disease was a leading cause of mortality in 2014
A
Describes the patterns of leading causes of death in the United States in 1900 and 2014 according to the guest lecture
4
Q
According to the graph in the guest lecture describing life expectancy at birth and at 65 years of age, shown by race and sex, for the years 1900, 1950 and 2014
A
In comparison to white females, black females have fewer years of life remaining in 1900, 1950, and 2014
5
Q
Wearing a seatbelt is which type of prevention level?
A
Primary prevention
6
Q
Routine testing of the stool for occult blood with the hopes of detecting colon cancer is an example of
A
secondary prevention
7
Q
In order to disprove the miasmatic theory of disease, a theory that described the causes of a cholera epidemic in England in the mid 19th Century, what did John Snow do?
A
He went from house to house counting all deaths from cholera in each house, and determined which company supplied water to each home. He determined that houses that drank water from one company had higher mortality rates than those who used the other
8
Q
A woman has a history of breast cancer, and this cancer has spread to other parts of her body. She therefore goes regularly to her physician for treatments to extend her life and quality of life. This approach to prevention is:
A
Tertiary approach
9
Q
True or False
Ignaz Semmelweis argued that hand washing would prevent the spread of disease
A
True
10
Q
“_ is the study of how disease is distributed in populations and the factors that influence or determine this distribution” (Gordis)
A
Epidemiology
11
Q
“_ is the study of the distribution and
determinants of health-related states or events in specified populations and the application of this study for the control of health problems”
A
Epidemiology (Dictionary of Epidemiology)
12
Q
Levels of prevention:
To prevent disease before it develops so as to maintain health
A
primary
13
Q
Levels of prevention:
To diagnose and treat disease in its early stages so as to restore or improve health
- Often a subclinical diagnosis
A
Secondary
14
Q
Levels of prevention:
To reduce complications of disease and improve functioning and quality of life where possible
- Often already have clinical symptoms
A
tertiary
15
Q
In biostatistics, what does a “parameter” refer to?
A
- There is a true value for a parameter (whether we know that value or not)
- Population parameters are usually unknowable
- A parameter is an attribute of a population
16
Q
In a randomized control trial that you perform, you conclude that a significant difference exists between your experimental group and the control group. What action can you take in relation to the null hypothesis?
A
Reject the null hypothesis
17
Q
- A _ is the probability, given that the null hypothesis is true, of obtaining a statistic as extreme or more extreme than the statistic you actually observed
- A _ of 0.0001 indicates your data are not very compatible with the null hypothesis
A
p-value
18
Q
You are doing a study to identify the mean cholesterol level of university students who eat at Panda Express every day. What will strongly affect your sample size:
A
Standard error of the student’s mean cholesterol level in your sample
19
Q
Confidence interval is statistically _ at the 0.05 level
- OR = 0.90 (95% CI = 0.85, 0.95)
- OR = 0.90 (95% CI = 0.83, 0.97)
A
significant
20
Q
An individual who harbors an organism but does not show overt clinical illness
A
The definition of a carrier according to class lecture
21
Q
Refers to a worldwide epidemic
A
pandemic
22
Q
Is defined as disease occurring rapidly and in excess of what is expected in a geographic region
A
epidemic
23
Q
Is defined as the habitual presence within a given geographic area
A
endemic
24
Q
What does the SIR model represent?
A
A model that represents how infectious agents are spread in the population
- Susceptible, Infected, Resistant
25
_ is defined as the resistance of a group of people to an attack by a disease because a large proportion of the members of the group are immune
Herd immunity
26
Epidemiologic triad consists of:
vector, host, environment, agent
27
Epidemiologic Triad:
- must be susceptible (for disease to occur), which is determined by a number of factors, including but not limited to:
- Biological factors (Genetic Profile, Family Background, Previous diseases, Immune status)
- Socio-Demographic characteristics (Religion, Age, Customs, Family background, Gender, Race, Socioeconomic Status, Education,Occupation, Social Inequalities)
host
28
Epidemiologic Triad:
- Biological
- Bacterium
- Virus
- Prions
- Chemical
- Physical
- Traumatic injury
- Radiation
- Nutritional
Agents
29
Epidemiologic Triad:
- Temperature
- humidity
- altitude
- housing/crowding
- neighborhood
- water/milk
- food
- air pollution
environment
30
Epidemiologic Triad:
- An organism that transmits a pathogen/disease
- Is not required for all diseases, but essential for some diseases
vector
31
Disease severity:
_ presents with signs and symptoms
Clinical disease
32
Disease severity:
Nonclinical disease can be:
- _: not yet clinically apparent but will be
- _: not clinically apparent and likely won’t be
- Preclinical
- Subclinical
33
Disease severity:
_: disease continues for years and may vary
with clinical manifestation(s)
- Includes diseases that may have later symptoms difficult to associate with original disease manifestations
Persistent (Chronic)
34
Disease severity:
_: no active multiplication of the agent and active signs or symptoms for some period of time
Latent period
35
The time interval from receipt of infection to
the time of onset of clinical illness
- May be affected by initial dose received
- In some diseases, individuals may be able to transmit the disease prior to showing clinical signs and symptoms of the disease
Incubation Period
36
_ is the act of generalizing from a sample to a population with calculated degree of certainty
Statistical inference
37
A _ is an attribute of a sample
- provides an estimate of a parameter
statistic
38
A _ is an attribute of a population
- Population _ are unknowable (usually)
- We believe there is a true value
parameter
39
We are curious about _ in the _
parameters in the population
40
We calculate _ in the _
statistics in the sample
41
A term for all of the new cases/person time at risk in the person-years graph figure
incidence density
42
If the number of new cases is known in a population, what is needed in order to calculate the incidence density (incidence rate)?
Total person-time
43
At the beginning of the school year, the prevalence of students who have chlamydia at UNR is 30%. At UNLV, the prevalence of chlamydia is 50%. Can we definitively conclude that cumulative incidences of chlamydia is higher at UNLV than at UNR?
No, because we do not know how long students have had chlamydia
44
Change lead to an increase in the prevalence, a decrease in the prevalence, or the prevalence remaining the same?
- A new measure is developed that prevents new cases of disease from occurring
decrease prevalence
45
Change lead to an increase in the prevalence, a decrease in the prevalence, or the prevalence remaining the same?
- There is immigration of a large number of healthy people into the population
decrease prevalence
46
Change lead to an increase in the prevalence, a decrease in the prevalence, or the prevalence remaining the same?
- A treatment is developed that prolongs the life of people suffering from the disease
increase prevalence
47
True or False
When the first AIDS cases were detected in the early 1980s, the case-fatality rate was high (over 90%). The disease duration was relatively short. Therefore the mortality rate could be used as an indicator of incidence
True
48
The number of diseased persons in a population at a specific time divided by the number of persons in the population at
that time
- A “snapshot” of disease occurring in a population
– Does not account for the duration of a disease and therefore is not a measure of risk
- is a useful measure of the burden of disease in a community (population)
- Can be useful for diseases that are difficult to determine the onset or beginning such as asthma or obesity
Prevalence
49
# of cases of a disease present in the population at a specified time / # of persons in the population at that specified time
prevalence
50
Disease duration is a major determinant of population _
prevalence
51
Where prevalent disease is _ of disease at a particular time, incident disease is _
- existing cases
- new disease cases
52
Incidence generally represented by two
measures:
– Cumulative incidence (or risk)
– Incidence rate (or incidence density)
53
The number of people who develop a
disease divided by the total number of people at risk of developing that disease over a specified period of time
- often referred to ask risk
- must describe the time period
cumulative incidence (risk)
54
of new cases of a disease occurring in a population during a specified period of time / # of persons who are at risk of developing the diseases during that period of time
cumulative incidence (risk)
55
IF incidence rates are not changing and in-migration is equal to out-migration then:
* Prevalence = _
Incidence x duration of disease
56
Prevalence depends on both _ and _
- incidence
- disease duration
57
Rate of new cases of a disease in a population at risk for the disease
- person-time
Incidence density (incidence rate)
58
of new cases of disease occurring in a population during a specified period of time / Total person time
Incidence density (incidence rate)
59
Used to describe quick outbreaks of disease (food poisoning example in previous in-class activity)
– Differs from IR in that time is not explicitly specified but is implicitly known
Attack rate
60
* Annual death rate = Total number of deaths in 1 year/Population at midyear
* Midyear population is an approximation, and by using it each person is contributing one person year (so it can be considered a rate)
* Can use a multiplier (e.g., multiplying per 1000, to get rate per 1000 p-y)
* Anyone in the population (denominator) must be able to convert to the numerator
Mortality rates
61
of deaths from a specific cause in one year / # of persons in the population at midyear
mortality rate
62
of individuals dying from a specific disease during a specified period of time after disease onset or diagnosis / # of individuals with the specified disease
case fatality rates (%)
63
Measure of the proportion of deaths (total) due to a specific disease or cause
* Not a rate but a proportion
* Changes in proportionate mortality over time may be due to changes in the mortality of another disease(s)
* Cannot tell us the risk of dying from a disease (it is not a risk or a rate)
proportionate mortality
64
Deaths caused by a specific disease / Total deaths in the population
proportionate mortality
65
The rate of disease in a population
morbidity
66
Death
mortality
67
Differences in age distribution between populations is very important to consider because age is the single most important predictor of mortality
- Can appropriately compare mortality between populations by _
- Direct Age _
- Utilize a “standard population” to calculate expected mortalities
from each age group of each population
- This method helps us to fairly compare older populations (e.g.,Florida) to younger populations (e.g., New York)
adjustments
68
A system by which a health jurisdiction receives reports submitted from hospitals, clinics, public units, and other resources
- inexpensive
passive surveillance
69
A system employing staff members to regularly contact health care providers or the population to seek information about health conditions
active surveillance
70
- Identified that cholera cases were originating from the Broad Street Pump in London
- Hypothesized the existence of the fecal-oral disease transmission route
Contributions John Snow made to the field of epidemiology
71
Type of prevention:
Physical therapy that is designed to relieve complication from advanced arthritis
tertiary
72
Type of prevention:
Administering Typhoid Vaccine to soldiers deploying to endemic areas
primary
73
Type of prevention:
Routine mammogram for detecting breast cancer
secondary
74
Disease measurement:
The percentage of disease-free UNR freshman who contract tuberculosis for their first time before the end of their freshman year
cumulative incidence
75
Disease measurement:
The number of new HIV cases diagnosed among injection drug users during 100 person years of follow-up
incidence rate
76
Disease measurement:
The percent of UNR seniors who currently have SARS-CoV-2 antibodies
prevalence
77
Disease measurement:
The proportion of people who were diagnosed with lymphoma cancer who died within 1 year of their diagnosis
case fatality rate
78
- _ ends at the onset of clinical illness
- During the entire _, an infected person will show no clinical signs of disease.
incubation period
79
Influence the prevalence of a disease in a population over time:
A new test is developed that increases the number of new cases of disease that are diagnosed
increase prevalence
80
Influence the prevalence of a disease in a population over time:
There is immigration of a large number of unhealthy people (most of whom are cases of the disease)
increase prevalence
81
Influence the prevalence of a disease in a population over time:
A treatment is developed that prolongs the life of people suffering from the disease
increase prevalence
82
Cases of illness are more likely to be missed in passive surveillance, while identification of cases using active surveillance are more likely to be complete
Difference between passive and active surveillance
83
Women who have had hysterectomies (removal of uterus) should not be included in incidence studies for uterine cancer, because they are not at risk of having the cancer. What would be the effect on incidence rates of uterine cancer if women with hysterectomies were included in the the denominator of the calculations?
The incidence rate would tend to decrease
84
True or False
In order for the incidence rate to be meaningful, any individual who is included in the denominator must have the potential to become part of the group that is counted as the numerator
True
85
_ follow exposed and unexposed individuals over time and ascertain who develops disease in each group
Cohort studies
86
You are a public health officer in Washoe County. A contagious illness has infected some of the inhabitants, and you do not have a laboratory test for the infection. You decide to place a quarantine on those showing any signs or symptoms of the disease. What is true?
The quarantine will not be effective if during the incubation period, those infected are not showing clinical symptoms but are still infectious
87
True or False
The ecological study compares group level measures across populations
True
88
Retrospective cohort studies are characterized by:
- The study groups are exposed and non exposed
- Incidence rates may be computed
- The required sample size is similar to that needed for a prospective cohort study
89
The exposure distribution among the cases (i.e., people with disease) is compared with the exposure distribution among the controls
True in a case-control study
90
True or False
Prospective cohort studies are more time-consuming and more expensive than retrospective cohort studies
True
91
Probability test is positive, given the individual has disease
* The fraction of the diseased who test positive
* Numerator: number of individuals who have disease and test positive (true positives)
* Denominator: number of individuals with disease (true positives+ false negatives)
* Ideal is 100%, i.e., all cases of disease are identified as such by the test
Sensitivity
92
Probability test is negative, given the individual does not have disease
* The fraction of the non-diseased who test negative
* Numerator: number of individuals who do not have disease and test negative
* Denominator: number of individuals without disease (true negatives +false positives)
* Ideal is 100%, i.e., all non-cases of disease are identified as such by the test
Specificity
93
Sensitivity =
a / a+c
94
specificity =
d / b+d
95
Probability individual truly has disease, given
a positive test
* The fraction of the positive tests who truly have disease
Positive Predictive Value (PPV)
96
Probability individual truly does not have
disease, given a negative test
Negative Predictive Value (NPV)
97
Negative Predictive Value (NPV) =
d /c+d
98
Positive Predictive Value (PPV) =
a / a+b
99
- A limitation of the design is the inability to determine the temporality of exposure and disease because they are both measured at the same time
- It is a design that measures incident disease
cohort studies
100
A risk difference greater than 0 implies a _ exposure
harmful
101
A risk difference less than 0 implies a _ exposure
protective
102
Children who received a vaccine had X times the risk of developing varicella when compared to those children who did not receive the vaccine
The correct interpretation for the risk ratio
103
What are measures of association?
- rate ratio
- risk ratio
- prevalence ratio
104
True or False
When comparing Cohort studies with Randomized Trials: Both studies compare groups that are exposed and unexposed. The difference between these two designs is the absence or presence of randomization
True
105
A population of 1000 children were followed from birth until age 10 years. 112 of the children developed asthma over the 10-year period. What was the odds of developing asthma over the 10-year period?
112/888 = 0.13
106
Malignant mesothelioma is a rare cancer that develops from the cells of the mesothelium, the protective lining that covers many of the internal organs of the body. The latency period is 35 to 40 years between exposure and diagnosis. Which study design would you recommend to investigate the predictors of this illness?
case control study
107
Formula for Risk Ratio
(A/A+B)/(C/C+D)
108
Correct Interpretation of Relative Risk/Risk Ratio and Odds Ratio:
If the Relative Risk/Risk Ratio or Odds Ratio is = 1
The risk is equal and there's no evidence of association
109
Correct Interpretation of Relative Risk/Risk Ratio and Odds Ratio:
If the Relative Risk/Risk Ratio or Odds Ratio is > 1
The risk in the exposed group is greater, association is positive, and exposure is possibly harmful for the disease
110
Correct Interpretation of Relative Risk/Risk Ratio and Odds Ratio:
If the Relative Risk/Risk Ratio or Odds Ratio < 1
The risk in the exposed group is less, association is negative, and exposure is probably protective
111
Correct formula for Odds Ratio
(A*D)/(B*C)
112
- It is an absolute measure of association
- It can have a negative or positive value
- It cannot be calculated from a case control study
risk difference
113
Relative risk cannot be calculated because risk is not measured in case-control studies
True regarding case-control studies and relative risk ratios
114
True or False:
Two common limitations of cross-sectional studies are:
1. It can be difficult to determine which came first, the exposure or the disease, when they are measured simultaneously
2. The design only provides a measure of disease prevalence, which is affected by both incidence and duration of disease
True
115
Strategies for reducing random error, selection bias, or information bias:
Increasing the number of people in your study
random error
116
Strategies for reducing random error, selection bias, or information bias:
Minimizing loss to follow up
selection bias
117
Strategies for reducing random error, selection bias, or information bias:
Measuring the outcome in both exposure groups identically
information bias
118
Strategies for reducing random error, selection bias, or information bias:
Measuring the exposure accurately
information bias
119
True or False
The width of confidence intervals (precision), statistical tests, and p-values are used to address systematic error (bias) but not random error
False
120
True or False
In order to have external validity, a study must have internal validity
True
121
A crude measure of association (e.g., OR, RR) between an exposure and an outcome can be biased by confounding. In what direction can the estimated measure be biased?
- It can be biased downwards (lower than truth)
- It can be biased upwards (higher than truth)
- It can be masked (ie., show no effect when there really is one)
122
True or False
The exposed and unexposed are not exchangeable when confounding is present.
True
123
Identify which variable is the exposure, which is the outcome, and which is the confounder:
- Sex = _
- Work location = _
- malaria = _
- exposure
- confounder
- outcome
124
True or False
Whereas screening tests are conducted on people who have not manifested any signs of disease, diagnostic tests are performed on individuals suspected of having the disease.
True
125
Dr. Silva is working at a pregnancy clinic in Rio de Janeiro, Brazil and because of increased risk of microcephaly due to Zika virus infection, she tests all of her patients for Zika infection using a blood test. Dr. Garcia is working at a pregnancy clinic in Miami, FL, USA and also tests all her patients for Zika virus infection using the same blood test. Zika virus infection is much more common in Rio de Janeiro, Brazil than Miami, Florida. Which of the following measures will be the same and which will be different between the two locations.
sensitivity = _
Specificity = _
Positive predictive value = _
Negative predictive value = _
sensitivity = same
Specificity = same
Positive predictive value = different
Negative predictive value = different
126
n screening for a disease such as Ebola, where a false negative result is highly undesirable both for the individual and for his/her contacts, public health professionals often desire a test that maximizes:
sensitivity
127
measures of association in the list have a null value of 0 (no evidence of association)
- risk difference
- prevalence difference
- rate difference
128
A study that examines conditions and events that already occurred or will occur anyway
Observational study
129
Study in which conditions are under direct control of the investigator
experimental study
130
The cross sectional, cohort, case control and
experimental studies reviewed all use an
_ person as the unit of analysis
Individual
131
Ecological studies use _ as the unit of analysis
groups or populations
132
What kind of study?
Researchers wanted to examine and identify risk factors for suicide attempts among patients with psychiatric disorders. The study enrolled 146 subjects who had attempted suicide and matched 146 subjects in the comparison group who had not attempted suicide. The researchers then compared sociodemographic characteristics between the cases and the controls. The researchers found suicide
attempt rates were higher in cases with certain sociodemographic characteristics
Case Control Study
133
What kind of study?
A study was conducted that investigated alcohol consumption and its relationship to problem gambling among young adults. All subjects were given a survey in which they were asked about the amount of alcohol they consumed as well as
gambling behavior simultaneously. All collected data was self reported. It was determined that those adolescents that consumed alcohol had a significantly
greater risk of gambling related problems.
Cross Sectional Study
134
What kind of study?
A study was conducted that examined the association between physical activity and pancreatic cancer risk. Physical activity and other information were collected at the baseline interview of 72,451 women and 60,037 men. Participants were followed up 10 years through annual linkage with cancer registry in combination with in-person interviews taking place every 2-4
years. At the end of the study 225 female and 159 male cases were identified. Adult exercise and physical activity were significantly associated with a decreased pancreatic cancer risk in men but not in women.
Prospective Cohort Study
134
What kind of study?
A study examined the links between diet and alzheimer's disease. The study compiled the prevalence of alzheimer’s disease in 11 countries along with dietary supply factors. The study found that counties with high total fat and total energy (caloric) supply were strongly correlated (i.e., risk factor) with high prevalence of alzheimer’s. While, countries with high fish and cereal/grain consumption were inversely correlated (i.e., protective) with
low alzheimer’s prevalence.
Ecological Study
135
What kind of study?
Researchers wanted to investigate the efficacy of a new drug designed to
treat prostate cancer. 958 men suffering with prostate cancer were recruited
for the study. Half of the subjects were randomly assigned to the new
treatment group while the other half were treated with Abiraterone Acetate,
a drug currently used to treat prostate cancer. It was found that the new
drug was not as effective as the currently used drug.
Randomized Controlled Trial
136
What kind of study?
A group of individuals working at a chemical plant had their medical records evaluated and were found to have high rates of thyroid, breast, and prostate cancer. Researchers examined their employee records from the last 30 years to determine the employee’s exposure to certain chemicals. The researchers found employees with high levels of exposure to a certain chemical had higher incidence of cancer compared to other employees.
Retrospective Cohort
137
_ is a group of people (population) that has certain (many) characteristics in common
Cohort
138
- Identify a study population of disease-free individuals (“at risk” population) to sample for study
- Identify who is exposed and unexposed to your exposure of interest
- Two groups of participants (exposed and unexposed) are ollowed over time
cohort studies
139
_ compare the incidence (risk or rate) of
disease of the exposed to unexposed groups to determine if they are different
- Comparison of the incidence rates is a statistical analysis
Cohort studies
140
- Measures incidence of disease (risk, rates)
- Good for studying rare exposures
- Can sample group of people getting the exposure of interest (e.g., gastric bypass surgery)
- Clear temporal relationship between exposure and outcome
– i.e., we know the exposure came before the outcome
- It is possible to evaluate multiple exposures and multiple diseases in the same _
Advantages of cohort studies
- cohort study
141
- Inefficient for studying rare diseases
– Requires large sample sizes
- Can be expensive to follow people over time
- Some Potential Biases in Cohort Studies
– Information bias
– Bias from nonresponse and loss to follow up
Disadvantages of cohort studies
142
Investigator moves through time with study
– Investigator can decide what data to collect during study because follow-up time has not yet occurred
– Can take a long time (have to wait for disease to develop)
– Expensive
Prospective (concurrent)
143
– Rely on historical records (not necessarily a problem)
- Limited to what information was recorded - investigator
cannot go back in time to add additional data collection
– Efficient for diseases with long induction and latent period
– Takes fewer resources
Retrospective (non-concurrent or historical cohort study)
144
A _ is a cohort study where individuals are randomized to be exposed or unexposed
randomized trial
145
Randomization is the _ because it is expected to balance the risk factors for disease in both the exposed group and the unexposed group
“gold standard”
146
When two groups of people are not exchangeable we say there is _
confounding
147
_ don’t have this same expectation of exchangeability
- In an _ we have to assume or hope the exposed and unexposed are exchangeable
- It can be hard to know if our assumption is correct.
- This property is why randomized trials are considered to provide “stronger evidence” than _
observational studies
148
Randomization is expected to result in _
exchangeability
149
Whenever possible, both the patient and the medical provider should be unaware of the treatment assignments
“double blind” study
150
- Overarching purpose: bias reduction
- Patient _ (e.g., with a placebo)
– Assures that subjects adhere similarly post-treatment
* Provider _
– Avoids subjective assessment and decisions
blinding
151
Pexp / Punexp = (a / (a+b)) / (c / (c+d))
Prevalence Ratio
152
CIexp / CIunexp = (a / (a+b)) / (c / (c+d))
Risk Ratio (Cumulative Incidence Ratio)
153
IRexp / IRunexp = (a / PTexp) / (c / Ptunexp)
Rate Ratio
154
The number of times an event occurs divided by the number of times it does not occur
Odds ratio
155
oddsexp / oddsunexp = (a / b) / (c / d) = (a*d)/(b*c)=AD/BC
Odds Ratio
156
- A ratio of two measures of
disease frequency
- Gives information about the strength of the
association
Relative Measures
157
- The difference between two measures of disease frequency
- Give Info about public health impact of exposure
- When baseline risk is really low, absolute measures may be more telling
Absolute Measures
158
- Randomized control trials
- prospective & retrospective cohorts
incidence studies
159
- cross-sectional
prevalence studies
160
Prevalence difference =
(A/A+B)-(C/C+D)
161
prevalence ratio =
(A/A+B)/(C/C+D)
162
odds ratio =
(A/B)/(C/D)=A*D/B*C
163
Risk difference =
(A/A+B)-(C/C+D)
164
risk ratio =
(A/A+B)/(C/C+D)
165
rate difference =
(A/person-time exposed)-(C/person-time unexposed)
166
rate ratio =
(A/person-time exposed)/(C/person-time unexposed)
167
interpret a rate ratio
The outcome rate among the exposed is X times the rate in the unexposed
168
interpret a risk ratio
The outcome risk among the exposed is X times the risk in the unexposed
169
interpret an odds ratio
- Disease odds ratio: Those with the exposure have OR times the odds of the outcome compared to the unexposed
- Exposure odds ratio: Those with the outcome have OR times the odds of exposure compared to those without the exposure
170
- Different samples will have slightly different effect estimates due to chance
- No consistent change from the universal truth across multiple sample, thus it averages out
Random error
171
- Factors that consistently affect measurement of an exposure or outcome or selection of participants across a sample
- Bias can be positive or negative but it is consistent and directly affects what is measured
- The bias doesn’t average out as sample size gets bigger
Systematic error/bias
172
“Systematic error that results in an incorrect or invalid estimate of the measure of association”
- Is primarily introduced by the investigator or study participants
Bias
173
Bias
- Can create spurious association when there really is none _
- Can mask an association when there really is one _
- bias away from the null or overestimate of effect
- bias towards the null or underestimate of effect
174
Systematic error in the recruitment or retention of study participants that distorts the measure of association between the exposure and outcome
selection bias
175
Types of selection bias
1. Loss to follow up
2. Exclusion bias
3. Sampling bias
4. Participation bias
176
Type of selection bias:
occurs when those who don’t participate over the course of the study are systematically different from those who do
-Example – move, bored, too ill, die, family responsibilities
Loss to follow up
177
Type of selection bias:
different eligibility rules are applied to cases and controls
- Example – allow those who drink alcohol in the cases but not in the controls
Exclusion bias
178
Type of selection bias:
occurs due to using a non-random sample
of a population
- Using a computer to collect data may exclude those whocan’t afford a computer
Sampling bias
179
Type of selection bias:
occurs due to different rates of participation/refusal across study groups
Participation bias
180
Occurs after subjects are selected into the study
- Due to flawed data collection or data entry or improper definition of variables
- Results in incorrect classification of exposed / unexposed or diseased / non-diseased (misclassification)
information bias
181
Types of information bias:
1. recall bias
2. reporting bias
3. interview bias
4. surveillance bias
182
Type of information bias:
Accuracy in recall of information (e.g., exposure) differs for each group (e.g., cases and controls)
recall bias
183
Type of information bias:
selectively revealing or suppressing of
information about past medical history
- Typically on sensitive issues like illicit drugs, STIs, alcohol, smoking
reporting bias
184
Type of information bias:
Systematic difference in solicitation, recording, or interpreting of information by the interviewer.
- Can occur when exposure information is sought when outcome is known (e.g., case-control), or when outcome information is sought when exposure is known (e.g., cohort
study or RCT).
interview bias
185
Type of information bias:
Patients in one exposure group have a higher probability of having the study outcome detected, due to increased surveillance, screening or testing of the outcome itself
surveillance bias
186
- Design studies appropriately
- Use the best study design as feasibly possible
- Use blinding
- Use randomization
- Maximize response rates and follow up rates
- Use validated and pre-tested data collection tools
- Train interviewers and observers
Employ consistent methods across groups
- Acknowledge and address limitations
Minimize bias
187
What study design:
A study compares country-level mean height across populations
ecological study
188
What study design:
Is the incidence rate of STI is lower among women who exercise compared to women who do not exercise on a regular basis
cohort study
189
What study design:
For men diagnosed with prostate cancer, is 10-year survival improved for those allocated to be treated with “Apalutamide” in addition
to surgery and chemotherapy, as compared to control men treated with only surgery, chemotherapy and a placebo drug
randomized control trial
190
What study design:
Are men with diabetes more likely than a group of hospital controls without diabetes to have a history of owning drinking diet beverages
case-control study
191
- The exposure distribution among the cases is compared with the exposure distribution among the controls
- You are over-sampling disease
- It is an appropriate study design for rare diseases
case-control studies
192
True or False
The underlying risk of a population can be calculated in a case-control study
false
193
Randomly allocate people into two different arms (increases exchangeability between exposed and unexposed; reduces confounding)
Randomization
194
Randomly select people from the population to be in the study (reduces selection bias)
Random sample
195
Relative measures (dividing)
– Risk Ratio
– Rate Ratio
– Prevalence Ratio
– Odds Ratio (OR)
196
Absolute measures of association (subtracting)
- Risk difference (RD)
– Rate difference (RD)
– Can do a prevalence difference (less common)
– Cannot ever do an odds difference
197
Measures of disease frequency
- risk
- rate
- prevalence
- odds
198
Measures of association
- risk ratio
- rate ratio
- prevalence ratio
- odds ratio
- risk difference
- rate difference
- prevalence difference
199
– Temporality (only one time point)
– Disease duration (some people have diseases for a long time)
* Remember: prevalence is affected by both incidence of disease and duration of disease
Major limitations of cross-sectional studies
200
– Information bias
* Recall bias
* Surveillance bias
* Interviewer bias
* Reporting bias
– Selection bias
* Loss to follow-up bias
* Sampling bias
* Participation/nonresponse/refusal bias
Systematic error
201
– Statistics quantify random error from sampling
– Interpret statistics
Random error
202
Random error:
_ allow us to assess whether a result observed, such as a risk ratio, is unlikely under the null hypothesis (null RR=1, null RD=0)
– Evaluating the role of chance/random variation
– NOT bias!!!!
Statistical tests
203
As sample size increase, the confidence interval will get more _
narrow
204
As sample size increases, the standard error _
decreases
205
Type of bias example:
“but authors were concerned as their
measure relied on self-reports for the outcome measure”
information bias
206
nfection with Bacillus daltonico is diagnosed by identifying green bacterial
colonies on a red agar gel. In a cohort study investigating inpatient infection with
B. daltonico as the outcome, the lead investigator discovers that one of the lab
technicians is red-green color blind. What would your biggest concern be with
respect to the validity of this study?
information bias
207
Loss to follow up is likely to create which kind of bias?
selection bias
