Epi Final

Front 1

Numerator

Back 1

The number above the line in a common fraction showing how many of the parts indicated by the denominator are taken, for example, 2 in 2/3.

Front 2

Denominator

Back 2

The number below the line in a common fraction; a divisor. A figure representing the total population in terms of which statistical values are expressed.

Front 3

Direct Adjustment

Back 3

In direct age-adjustment, a common age-structured population is used as standard. This population may actually exist (e.g., United States population, 1999) or may be fictitious (e.g., two populations may be combined to create a standard).

Front 4

Epi

Demos

Logos

Back 4

upon

people

thestudy of

Front 5

Epidemiology

Back 5

Study of distribution and determinants of health related states or events in specified populations and the application of this study to control of health problems

Front 6

What are some goals of epidemiology?

Back 6

Determine extent of disease in community

Study natural history and progression of disease

Provide foundation for developing public policy

Front 7

What are some CDC recognized health achievements?

Back 7

safer workplaces, vehicle safety, cardiovascular disease prevention, public health preparedness - more infectious diseases in global risk

Front 8

What work remains to be done according to the CDC?

Back 8

health behaviors, drug resistance, mental health, poverty, safety of the food and water supply

Front 9

What are the three types of prevention?

Back 9

1) Primary: prevent initial development of the disease (sun screen)

2) Secondary: early detection of existing disease (screening)

3) Tertiary: reduce the impact of the disease (medication)

Front 10

Non-modifiable

Back 10

age, male or female

Front 11

Modifiable

Back 11

lifestylechoices

Front 12

Populationbased approach to prevention

Back 12

inexpensive, can't be too invasive or too specific, ex) anti-smoking campaign

Front 13

High-riskapproach to prevention

Back 13

target sub-groups - this is more costly and invasive

Front 14

What is the clinical practice of Epidemiology?

Back 14

1) Diagnosis = how do we know to make a clinical diagnosis =population level data

2) Prognosis = observe large populations = how do you know the fluwill end in 48 hours

3) Selection of therapy = clinical trials = observations made byepidemiology study designs

Front 15

What are the three core factors of descriptive epidemiology?

Back 15

1) place

2) time

3) person

Front 16

What are the four types of measures?

Back 16

1) Absolute: total counts. ex) N=5 persons admitted for TB 1967-1969

2) Proportions: numerator divided by the denominator (not a rate). Part over whole, part is in whole. ex) Estimateof baseline condom use last sex = 49/175 = 0.28 = 28%

3) Rate: number of cases divided by the population at risk - must include time. ex) There were 60 cases per 100,000 population in 1987

4) Ratio: numerator divided by the the denominator. Numerators is not in the denominator. ex) for every 1 female there were 7.2 males

Front 17

Morbidity

Back 17

The condition of being diseased

Front 18

Mortality

Back 18

A countable death

Front 19

What is the pyramid from healthy to diseased states?

Back 19

People treated

People diagnosed

People Affected

People at risk

Total People

Front 20

Progression from healthy to diseased state

Back 20

Total people -> sick -> sick who seek care -> sick hospitalized

Front 21

What is the difference between a rate and a proportion?

Back 21

A rate is how fast the disease is occurring in the population. A proportion is what fraction of the population is affected.

Front 22

What is a fixed or closed population and what formula is used?

Back 22

Membership is permanent and/or defined by an event. Something that occurred during Katrina = the people that remained in the city during Katrina defined by an event. If no one is lost from a study.

Use cumulative incidence rate formula

Front 23

What is a dynamic or open population and what formula is used?

Back 23

Membershipis transient and defined by being in orout of a specified state. Ex)Residents of DC. Students at GW.

Use incidence rate formula

Front 24

Incident

Back 24

Incidence measures characterize new cases of a disease in a specified time frame among persons at risk of the disease

Front 25

Prevalent

Back 25

Prevalent cases are those whose disease developed or was diagnosed before they were identified for the study.

Front 26

Cumulative incidence

Back 26

The probability that a particular event, such as occurrence of a particular disease, has occurred before a given time.

Front 27

What are some key things to remember about incidence?

Back 27

- Denominator and numerator must have same time frame

- Everyone in denominator must be at risk

- Numerator must be represented in the denominator

- People in the denominator must have been followed for the duration of the specified study period

Front 28

What are the two types of denominators for incidence when calculating morbidity measures?

Back 28

1) people at risk are observed through a defined time period (cumulative incidence)

2) people at risk are observed for the full time period (incidence rate - need person time)

Front 29

Point prevalence

Back 29

quantifies the number of existing cases of a disease in a population at one point in time

Front 30

What is person time and how do you figure it out?

Back 30

Person-time is an estimate of the actual time-at-risk – in years,months, or days – that all participants contributed to a study.

Front 31

Attack Rate

Back 31

The number of people exposed to a suspect food who became ill/the number of people exposed to that food (a proportion)

Front 32

Period prevalence

Back 32

quantifies the number of existing cases of a disease in a population over a specific period of time

Front 33

What kind of measures are prevalence measures?

Back 33

proportions not rates

Front 34

What is the spicket analogy?

Back 34

You have a spicket that is putting water into a bucket. The water is the incidence. The water contains old and new water i.e. old and new cases of the disease, which is the prevalence. To make sure the water doesn't overflow you need have deaths or cures, which is duration.

Prevalence = new cases * duration of illness

Duration = prevalence / incidence

Front 35

What is the relationship between incidence and prevalence?

Back 35

The thing happens (measure of risk) = incidence vs. the thing happens in a population at a specific point in time (i.e. the overall disease burden) = prevalence

Front 36

How would you calculate if someone has ever had asthma?

Back 36

Cumulative Incidence

Front 37

How would you calculate if a population has asthma?

Back 37

Point prevalence

Front 38

How would you calculate if a population had asthma in the last year?

Back 38

Period prevalence

Front 39

Mortality rate

Back 39

Total number of deaths / population at mid-year

Denominator is entire population at risk of dying from disease (includes those who have and who do not have the disease)

Front 40

Case fatality

Back 40

The case-fatality rate is the proportion of persons with a particular condition (cases) who die from that condition. It is a measure of the severity of the condition.

·Denominatoris limited to those who have the disease ·Measure of severity of disease ·Not really a rate but a proportion

Front 41

Proportionate mortality

Back 41

The proportion of all deaths due to a specific cause. Proportionate mortality describes the proportion of deaths in a specified population over a period of time attributable to different causes. Each cause is expressed as a percentage of all deaths, and the sum of the causes must add to 100%. These proportions are not mortality rates, because the denominator is all deaths rather than the population in which the deaths occurred.

Front 42

How do you adjust for age - what two key things do you need?

Back 42

1) ensure you're looking at comparable numbers i.e. not absolute numbers

2) standardize against a norm - you're looking at age specific rates vs. all ages

Front 43

What is direct standardization?

Back 43

When age-specific mortality rates for two or more populations are known, direct standardization method can be applied.

You need: 1) age specific death rates for each age group 2) the number of persons in each age group

Front 44

What is indirect standardization?

Back 44

When Age-specific mortality rates of the population (s) of interest are unknown, indirect standardization method is applied.

Use when 1) age specific mortality rates are not available 2) the numbers of deaths are too small insome age groups in the groups to be standardized 3) you want to compare one population to astandard population

Front 45

Crude

Back 45

Crude rate is an overall average rate of disease, but it doesn't take into account possible confounding factors

- Benefit: Summary rate and easy to calculate

- Con: Differences may be due to other factors andnot true variation in rate (age, sex, race, etc) of population

Front 46

Adjusted

Back 46

Summary rate accounts for differing distributions, however,these are not useful for planning because we are comparing, but not treating everyone

- Benefit: Summary rate; accounts for differingdistributions of other factors so useful for comparing - Con: Not real rates so not as useful for planning;dependent on population used to standardize

Front 47

Specific

Back 47

A mortality rate limited to a particular group

Benefit: Homogenous subgroups; useful for planning Con: Difficult if there are multiple subgroups tocompare

Front 48

What can give you risk - incidence or prevalence - why?

Back 48

Incidence can give you risk. Prevalence cannot give you risk. Prevalence is only a proportion.

Front 49

Endemic

Back 49

Expected number of cases in a population

Front 50

Epidemic

Back 50

Greater number of expected number of cases in a population

Front 51

Outbreak

Back 51

Usually smaller scale and localized but used interchangeably with epidemic

Front 52

Cluster

Back 52

Unusual number of cases reported in given time or place-don’t always know what is expected

Front 53

Pandemic

Back 53

Global epidemic

Front 54

What is the epidemiological triad?

Back 54

Agent -> (Sometimes Vector) -> Host -> Environment

Front 55

Intrinsic vs. Extrinsic

Back 55

Factors relevant to the host - intrinsic

Factors relevant to the environment - extrinsic

Front 56

What are four ways infectious agents are classified?

Back 56

1) MicrobiologicProperties - bacterial

2) Mode of Transmission - food

3) PathogenicMechanism - fever

4) Reservoirof Organism - animal

Front 57

Infectivity

Back 57

Ability of agent to invade and multiply in host

Front 58

Pathogenicity

Back 58

Ability of agent to produce clinically apparent illness in host

Front 59

Virulence

Back 59

Ability of agent to produce severe disease in a host

Front 60

What is the iceberg concept?

Back 60

Most infections are, however, are detected only when body fluids or other samples become available for recognition of past exposure.

Front 61

Clinical

Back 61

Characterized by signs and symptoms

Front 62

Preclinical

Back 62

Not clinically apparent but destined to progress

Front 63

Subclinical

Back 63

Not clinically apparent but not destined to progress

Front 64

Persistent (chronic)

Back 64

fails to get rid of infection and it persists

Front 65

Latent

Back 65

Infection with no active multiplication of agent

Front 66

Carrier

Back 66

Individual who harbors organism but shows no evidence of clinical illness or antibody response

Front 67

Vaccine

Back 67

A vaccine is any biologically derived substance that elicits a protective immune response when administered to a susceptible host.

Front 68

What are the characteristics of an ideal vaccine?

Back 68

Illicit an immune response with lasting prevention, minimum adverse effects, protect across strains, easily administered

Front 69

What is herd immunity?

Back 69

“Resistanceof a group of people to an attack by a disease to which a large proportion ofthe group is immune.” (Gordis)

When the population is immune (immunized or infected) above the critical level (herd immunity level), the infectious disease will not be spread.

Front 70

What are the conditions for Herd Immunity?

Back 70

Disease agent restricted to single host with direct transmission

Infections must induce solid immunity (not partial)

Operates optimally when there is random mixing in population

Front 71

What are the characteristics of an outbreak investigation?

Back 71

Purpose:

◦Determine what is causing the disease ◦Prevent the spread of disease

Steps in an Outbreak Investigation:

◦Define the outbreak and validate its existence ◦Examine the distribution of cases by place and time

◦Look for interactions of relevant variables ◦Develop hypotheses

◦Testhypotheses

◦Recommend control measures

◦Prepare a written report

◦Communicate findings

Front 72

Point Source

Back 72

Single exposures as often point source - explosive, sudden,rapid increase, cases limited to people who share common explorer, food born outbreak

Front 73

Intermittent common source

Back 73

The epidemic curve of an intermittent common-source outbreak often has a pattern reflecting the intermittent nature of the exposure.

Front 74

Continuous common source

Back 74

High levels of illness throughout

Front 75

Propagated

Back 75

Measles outbreak at Disney - exposure, sick, people coming in contact with the sick. We can get the incubation period this way - go from first, to last case or fist to peak.

Front 76

What is a cohort study?

Back 76

- Exposed -> either the disease develops or not

- Not exposed -> either the disease develops or it does not

Front 77

Is a cohort study experimental or observational?

Back 77

It's in between

Front 78

What are we looking for in a cohort study?

Back 78

An association between a factor and a health outcome. The next step would be to derive inferences around a causal relationship.

Front 79

What is a prospective cohort study?

Back 79

Investigators at the beginning: Follow participants forward to determine the relative risk. The exposure came before the illness.

Front 80

What is a retrospective cohort study?

Back 80

Investigator at the end: Follow participants who have had both the exposure and outcome.

Front 81

What is an ambidirectional cohort study?

Back 81

Investigator in the middle: we have records of who was exposed and now we will follow them forward - if there is an environmental exposure its possible we didn't have a registry set up immediately because people may not have known it was a risk

Front 82

What three things do all cohort study designs have?

Back 82

1) participants categorized according to exposure

2) participants were disease free at study starting point

3) the exposure occurred before the outcome

Front 83

When are cohort studies useful?

Back 83

- rare disease

- natural history of disease

- confirm source of an outbreak

Front 84

What are the advantages of a prospective cohort study?

Back 84

◦Able tocollect contemporaneous data specifically for the study

•More expensive,time consuming

•Not efficient for diseases with longlatent periods

•Better exposure and other risk factor data

•Less vulnerable to bias(we’ll talk about this later)

Front 85

What are the advantages of a retrospective cohort study?

Back 85

◦Retrospectivestudies generally rely heavily on records

◦Cheaper,faster

◦Efficient with diseases with long latent period◦Exposure data may be inadequate

Front 86

How do you conduct a cohort study?

Back 86

1.Identifya group or cohort based on exposure or characteristic

2.Choose unexposed population for comparison 3.Follow groups forward in time

4.Determinedisease or mortality status of eachindividual in the cohort

5.Comparerates of disease or mortality in exposedgroup to rates in unexposed group

Front 87

What are external comparison groups in a cohort study?

Back 87

◦Standardized Mortality Ratio (SMR) ◦Standardized Incidence Ratio (SIR) and Standardized Rate Ratio (SRR)

Front 88

What is relative risk?

Back 88

Relative risk=risk of outcome among exposed / risk of outcome among unexposed

Front 89

How do you interpret relative risk?

Back 89

- If the RR=1.0, then the risk among exposed is the same as that among unexposed

- If the RR is greater than 1.0 then the risk among exposed is greater than that among unexposed - - If the RR is less than 1.0 then the risk among exposed is less than that among unexposed

Front 90

What is a life table?

Back 90

If we start loosing people throughout the study we use life tables - we have a number at risk, but we are losing people. Take the total number at risk. How many died over at risk. Who did not die - if 47% died 53% survived. Who survived from the beginning and who survived as we lose people - we are making assumptions about who we are losing and we are able to calculate cumulative

Front 91

Case Control Study

Back 91

Were exposed - Were not exposed - Have the Disease

Were exposed - Were not exposed - Do Not Have the Disease

Front 92

Are case control studies more observational or experimental?

Back 92

Observational

Front 93

What kind of study? To determine the long-term effectiveness of influenza vaccines in elderly people, groups of vaccinated elderly and unvaccinated community-dwelling elderly were followed post-vaccination. The results suggest that the elderly who are vaccinated have a reduced risk of hospitalization for pneumonia or influenza.

Back 93

Prospective cohort

Front 94

Suppose investigators wanted to test the hypothesis that working with the chemicals involved in tire manufacturing increases the risk of cancer. Investigators used employee health and employment records collected over the past three decades as a source of exposure data. They determined outcome by consulting cancer registries.

Back 94

Retrospective cohort

Front 95

What are two key elements of a case control study?

Back 95

1) Classify participants based on outcome or disease of interest 2) Look back in time to assess antecedent exposures

Front 96

When are case control studies useful?

Back 96

- When exposure data are expensive or difficult to obtain

- When disease has long induction and/or latent period

- When thedisease is rare

- When little is known about the disease

- Whenunderlying population is dynamic

Front 97

What are cases in case control studies?

Back 97

numerators

Rate of disease in exposed

Rate of disease in unexposed

Front 98

Why is a case-control study is a more efficient form of a cohort study?

Back 98

- Cases are the same as those that would be included in a cohort study - Controls provide a fast and relatively inexpensive means of obtaining the exposure experience in the population that gave rise to the cases

Front 99

What is a control?

Back 99

A sample of the source population that gave rise to the cases. Must not have the outcome of interest

Front 100

What is the purpose of a control?

Back 100

To estimate the exposure distribution in the source population that produced the cases

Front 101

When are generalpopulation controls used in case control studies?

Back 101

- Most often used when cases are selected from a defined geographic population -Sources:random digit dialing, residence lists, drivers’ license records

- Advantages of general population controls Because of selection process, investigator is usually assured that they come from the same base population as the cases - Disadvantages: ◦ Time consuming ◦ Expensive ◦ Hard to contact and get cooperation ◦ May remember exposures differently than cases (recall issues)

Front 102

What are advantages and disadvantages of hospital case controls in a case control study?

Back 102

Advantages: Same selection factors that led cases to hospital led controls to hospital} Easily identifiable and accessible (so less expensive than population-based controls)

Disadvantages: Since hospital based controls are ill, they may not accurately represent the exposure history in the population that produced the cases

Front 103

What illnesses make good hospital controls in a case-control study?

Back 103

Those illnesses that have no relation to the risk factor(s) under study

Front 104

Name flawed controls

Back 104

Special control groups like friends, spouses, siblings, and deceased individuals.}These special controls are rarely used} Some cases are not able to nominate controls because they have few appropriate friends, are widowed, or are only or adopted children}Dead controls are tricky to use because they are more likely than living controls to smoke and drink

Front 105

What are keys to matching in a case-control study?

Back 105

One way to ensure that the distribution of cases and controls on other factors is similar. Selecting controls that are similar to cases on certain characteristics

Front 106

Why can’t we just use prevalent cases to increase the size of our study?

Back 106

Risk factors identified in prevalent cases are often more related to survival than actual development of disease. ◦If people die soon after diagnosis, they would not be represented in prevalent cases. ◦Risk factors associated with surviving cases might not be a characteristic of all cases but only of survivors.

Front 107

How do you conducta case-control study?

Back 107

- Define study population - Case definition, sampling - Control definition, ascertainment of controls - Data collection (blinding may be needed) ◦Maybe real time or existing data - Data cleaning - Analysis

Front 108

What is the definition of odds?

Back 108

The ratio of the probability of an event occurring to that of it not occurring

Front 109

Odds Ratio

Back 109

Odds ratio =

odds of exposure among cases=

a/c= ad/bc

odds of exposure among controls = b/d Just like the incidence rate ratio and cumulative incidence ratio, the odds ratio is a ratio measure of association.

Front 110

Interpret the odds ratio = 1.3 women, lupus, henna

Back 110

A woman with lupus had 1.3 times greater odds of henna exposure than a woman without lupus.

Front 111

How do you Interpretingthe odds ratio?

Back 111

- If the OR=1.0, then the odds of exposure among diseased the same as that among non-diseased - If the OR>1.0, then the odds of exposure among diseased the greater than that among non-diseased - If the OR is 0 we don’t really have a temporal association (unless nested in cohort or exposure recorded previously - We say our exposure is associated with the outcome

Front 112

What is the rare disease assumption?

Back 112

The rare disease assumption postulates that as the prevalence of the disease under study decreases, estimates of the OR derived from the 2 ×2 table will approximate the RR that would have been derived from a cohort study - Does this mean that the odds described in a 2 × 2 table from a case-control study are really risks? No! The condition affects less than 5-10% of the underlying population

Front 113

What are requirements for control?

Back 113

• Controls must be representative of individuals in the base population. • Cases must be representative of individuals in the base population with the disease of interest.

Front 114

Case-control studies assume several key items

Back 114

- There is a good way to find cases. - There is a good way to find controls - The controls we use must be those that would have been identified as cases in our study, had they been cases. This means that the controls are a subset of our population that actually gives rise to our cases. - All the exposure information required was systematically maintained at the time the exposure under study took place, and these data are of sufficient detail and quality to describe the exposure.

Front 115

Strengths of case control studies

Back 115

- Often only approach for rare diseases or outcomes - Can leverage multiple controls (e.g., 1:4 case:control ratio) - OR can approximate RR if rare outcomes - Multiple exposures can be assessed - Useful when long latent period prior to disease

Front 116

Limitations of case control studies

Back 116

- Notable to characterize risk, only odds/odds ratios - Incidence cannot be estimated - Recall bias and other biases (more later!) - Challenges in identifying suitable controls - Relies on self report if records not available - Data on exposure and other variables usually obtained by interview or survey (subjective) - Can look at only one disease/outcome

Front 117

What is a nested case-control study

Back 117

A nested case control (NCC) study is a variation of a case-control study in which each case is matched to one or more controls based on participant characteristics, e.g. age. In contrast, in a standard case-control study, a set of controls are selected without matching

Bottom line: There is an original cohort population free of disease -> some develop the disease -> the best remain disease free -> they turn into the cases and controls

Front 118

What is a case-cohort study

Back 118

Cases are defined as those participants of the cohort who developed the disease of interest, but controls are identified before the cases develop.

Bottom line: A random sample of the cohort, called the sub-cohort, is used as a comparison group for all cases that occur in the cohort.

Front 119

What are the advantages of a nested case-control study?

Back 119

- Usually more efficient than cohort studies- Don’t have to calculate exposures on entire cohort, just on cases and selected controls - Controls come from same population as cases, thereby minimizing biases introduced by selection of controls - Efficient– not all members of parent cohort require exposure assessment - Flexible– allows testing of hypotheses not anticipated when the cohort was drawn (at beginning) - Reduces selection issues –cases and controls sampled from same population - Reduces problems with information– risk factor exposure can be assessed with investigator blind to case status

Front 120

What are the disadvantages of a nested case-control study?

Back 120

- Reduces power (from parent cohort) because of reduced sample size

Front 121

What are the advantages of a case cohort study?

Back 121

- Efficient–not all members of parent cohort require exposure assessment - Flexible–allows testing hypotheses not anticipated when the cohort was drawn - Reduces selection issues - cases and non-cases sampled from same population - Reduced problems with information – risk factor exposure can be assessed with investigator blind to case status - Other advantages, as compared to nested case-control study design: The sub-cohort can be used to study multiple outcomes (don’t have to select multiple control groups)

Front 122

What are the disadvantages of a case cohort study?

Back 122

- Increased potential for information issues because - sub-cohort may have been established after baseline - exposure information collected at different times (e.g. potential for sample deterioration)

Front 123

Cohort-Key Points

Back 123

- Selected on exposure (risk factor) status - Followed forward for outcome (whether in real time or historically) - Can determine temporality - Can estimate risk or incidence - Choice of unexposed group is important (internal vs. external) - Efficient for rare exposures and multiple outcomes - Longer time and more expensive but less flaws than case control - Multiple outcomes can be examined (sometimes also multiple exposures if cohort is based on population at start instead of exposures)

Front 124

Nested case control vs. Case cohort

Back 124

- Both are within cohorts and can establish temporality because exposure established before disease - Nested selects controls from those who were disease free at the time the case developed (you have a population and people get sick and you match them to people who have not developed the disease) - Case cohort selects controls as random sample from baseline cohort (you have a population of healthy and sick people, you match people randomly whether or not they are sick)

Front 125

Key characteristics of Cross-Sectional Studies

Back 125

- Begins with a population base but does not follow individuals over time - Looks at prevalence of disease and/or exposure at one point in time (a population“snapshot”) - Persons in the snapshot are classified as diseased or non-diseased, exposed or unexposed - Slice in time of the population

Front 126

A cross sectional study measures prevalence - what are the downsides of this?

Back 126

- May not be representative of all cases - Excludes those who died before study was carried out - Includes cases who are still alive - Not possible to determine temporality

Front 127

Where is the investigator in a cross sectional study?

Back 127

In the middle

Front 128

How do you conduct a cross-sectional study?

Back 128

1. Identify target population and sample 2. Define variables 3. Obtain data on disease, exposure, and other variables on as many targeted individuals as possible 4. a) Can compare prevalence of exposure between diseased and non-diseased individuals (POR) b) Can compare prevalence of disease between exposed and non-exposed groups (PR)

Front 129

What is the Prevalence Odds Ratio?

Back 129

- Odds of exposure among diseased compared to non-diseased (exposure-odds ratio); or - Odds of disease among exposed compared to unexposed (disease-odds ratio) - Either yields the same OR, which is a prevalence odds ratio - Calculated in the same manner as the odds ratio

Front 130

The Prevalence Ratio (PR)

Back 130

- Interpretation similar to the relative risk in cohort studies but in terms of prevalence - Calculations are similar to RR

Front 131

Serial cross-sectional studies

Back 131

- Multiple“slices” over time - Not same people but same methods, populations - Allows evaluation of secular and other changes - Much easier than longitudinal data collection

Front 132

What are the strengths of a cross-sectional study?

Back 132

- Useful for hypothesis generating - Relatively inexpensive compared with other designs (but not “cheap”—this is highly dependent on what is being done) - Maybe relatively quick compared with other study designs - No follow up reduces resources needed for study - Generally acceptable and feasible - Useful to study morbidity or pre-clinical markers - Can collect detailed data on exposures and other variables - Do not have to know disease/exposure status ahead of time

Front 133

What are the limitations of a cross-sectional study?

Back 133

- Prevalence odds ratio and prevalence ratio are descriptive and useful, but not able to characterize risk - Poor recall, information issues, mis-classification; limitations highly dependent on type of study and data collection ◦ Large range of issues in cross-sectional studies ◦ Compare self-report survey vs. detailed biomarker collection - Challenges in determining temporality due to simultaneous outcome and risk factor variable collection color

Front 134

What is an ecological study?

Back 134

- Studies of group characteristics. The “units of analysis are populations or groups of people, rather than individuals.” (Last, 1995) - Correlation between disease rates and exposures are based upon average exposure levels and average disease rates. - Ecological studies do have comparison groups. However: NO data on individual exposures!

Front 135

What are the limitations of ecological studies?

Back 135

- Subject to numerous biases and limitations - Results more likely to be explained by other factors than exposure of interest than individual risk studies - Time trend studies are limited in that you can’t determine whether exposure preceded the outcome - Subject to the “ecological fallacy”

Front 136

What is the ecological fallacy?

Back 136

- Results when correlations found on a group level do not hold true for the individual members of these groups - We don’t know the link between exposure and disease among individuals within each group

Front 137

What are the strengths of ecological studies?

Back 137

- Availability of data on exposures and disease - Can be done quickly and with limited resources - Exposures may differ substantially between cities, states, and countries - Analysis is fairly simple - Great for hypothesis generating

Front 138

Descriptive versus analytic studies

Back 138

Descriptive

- Are not hypothesis-testing, but generate hypotheses - Focus on patterns of disease by person, place, and time - Usually have no comparison group - Are“observing” what has occurred naturally - Investigator does not manipulate exposure status of study participants

Analytic Observational

- Test a hypothesis - Usually include a control or comparison group or comparison to baseline status - Use a more complex design - Observe exposure and outcome - Do not manipulate or assign

Front 139

What are descriptive studies?

Back 139

- Case Reports - Case Series - Surveillance - Ecological - Cross-sectional

Front 140

What are observational studies?

Back 140

- Cohort - Case-Control - Cross-Sectional

Front 141

What are experimental studies?

Back 141

- Randomized Trials - Non-Randomized Trials

Front 142

How would you phrase this measure of association: Oddsratio>1 (Example: OR=2)

Back 142

Cases had 2 (times) greater odds of exposure than controls.

Front 143

How would you phrase this measure of association:Oddsratio<1 (Example: OR=0.4)

Back 143

Cases had 0.4 times the odds of exposure than controls. OR Cases had 60% lower odds of exposure than controls.

Front 144

How would you phrase this measure of association: POR>1(Example: POR=2)

Back 144

Prevalentcases had 2 (times) greater odds of exposure than controls.

Front 145

How would you phrase this measure of association: POR<1(Example: POR=0.4)

Back 145

Prevalent cases had 0.4 times the odds of exposure than controls OR Prevalent cases had 60% lower odds of exposure than controls.

Front 146

Observational - analytic studies

Back 146

Are hypothesis testing

Front 147

Observational - descriptive studies

Back 147

Are not hypothesis testing

Front 148

Experimental analytic studies

Back 148

Are always hypothesis testing

Front 149

What is the Randomized Controlled Trial Design?

Back 149

Define study population -> Assign exposure -> exposed vs. non-exposed -> outcome vs. no outcome

This is a prospective study and there is an exposed outcome vs. no outcome groups and a non-exposed outcome vs. no outcome

Front 150

What are the key attributes of a Randomized Controlled Trial Design?

Back 150

- Group membership assigned by researcher - Study participants must be followed forward in time - Must employ at least one treatment /intervention applied in a standard way to all subjects in that group - Must contain a control (referent) group

Front 151

Breakdown of definition of a Randomized Controlled Trial Design

Back 151

- Trial= an experiment - Randomized= the exposure (intervention) is randomly assigned to study subjects (“randomized” to treatment groups) - Controlled= there is a comparison group (control group) with which the exposed (intervention) group is compared

Front 152

Key questions to ask in experimental study questions

Back 152

Is it ethical and is it feasible?

Front 153

Four types of intervention in trials

Back 153

1) prevention trial - can we prevent the disease? 2) therapeutic trial - can we better manage the symptoms of those who have the disease - when we don't yet have a treatment 3) equivalency trials - we have a number of treatments in the market and we want to see if it works just as well as the existing treatments 4) Superiority trial - is the current treatment better than the standard of care?

Front 154

Randomized trials can be clinical or prevention trials - what does that mean?

Back 154

Clinical: Therapeutic: Aim to slow, stop or reverse disease and you would enroll people with disease/conditionUsuallya well-defined population

Prevention: Aim to prevent disease/condition therefore you would enroll people without the condition and focus on people at highest risk

Front 155

What do you need in a study population in a randomized trial?

Back 155

- Need study subjects who are well defined - Study subjects must have potential or be “at risk” for outcome - Potential for outcome must be such that there will be sufficient endpoints for analysis in the study period - Need study subjects likely to complete follow-up in terms of both:◦Good short term survival◦Good adherence

Front 156

What is randomization in a randomized trial?

Back 156

- Refers to the random “allocation of participants into intervention arms” (i.e.assignment to treatment groups) - Unit of randomization is usually the individual in randomized controlled clinical and prevention trials - Occurs after admittance to study - Randomization scheme must be “unpredictable” so no possibility of “gaming the system”

Front 157

What is the goal of randomization?

Back 157

- To achieve baseline comparability between compared groups on factors related to outcome(“knowable” and “unknowable”)

Front 158

What are the types of randomization?

Back 158

- Simple: Everyone has 50/50 chance of getting treatment or not

- Block: Randomize in blocks of n people so that every n people have a 50/50 chance of getting assigned

- Stratified: Divided into groups based on variables and then randomized within groups. Avoid imbalance of sex or age, however, we need to know what we have in each strata, which is hard to manage

Front 159

Efficacy

Back 159

Can the treatment work and how well under ideal conditions? (Not necessarily in the“real” world)

Front 160

Effectiveness

Back 160

Does the treatment work in the real world? By average clinicians and average patients?so-style-text

Front 161

Efficiency

Back 161

Is the treatment/intervention cost (in terms of resources and money) the most effective for the cost, or are is there minimal cost for maximum benefit?

Front 162

Blinding (masking)

Back 162

Refers to the practice of not letting those involved with a study know which treatment individual participants are receiving.

Goal:To reduce bias in the assessment/reporting of outcomes, which could lead to problems if knowledge of the treatment effected the perception or measurement of the outcome.

Front 163

Single Blinding

Back 163

Subjects blind to treatment

Front 164

Double Blinding

Back 164

Both the subjects and clinicians don't know - gold standard, controls for uneven distribution, minimizes study errors, however, establishes efficacy not effectiveness

Front 165

Triple Blinding

Back 165

The data analysts, investigators, and subjects are blind

Front 166

Ways to deal with non-compliance

Back 166

- Can conduct sensitivity (“what if”) analysis to determine effect on the results - Can look at results stratified by adherence (but be cautious of conclusions) - Intention-to-treat is best analysis. Why?

Front 167

What is Intention to treat?

Back 167

Analyze as if you followed through with the entire study (analyze as randomized)

Front 168

What are the three randomized trial variations?

Back 168

1) Parallel group= subjects stay in assigned group through end of follow-up. This is the classic design. 2) Cross-over =subjects switch assigned group, usually halfway through follow-up. There may be a “wash-out” period in between. Limited applicability by both intervention and outcome but subject can be their own control. 3) Factorial design= More than 2 intervention groups, often with multiple combinations possible. Subjects stay in assigned group to end of follow-up. Allows more complex analysis and comparisons of treatment effects.

Front 169

What are the ways to protect the subject in a randomized trial?

Back 169

- Informed consent - Use of standard therapy (as comparison therapy) , rather than placebo - Subject has option to discontinue intervention, or completely drop out of the trial - Physician may choose to take subject off study drug to place him/her on “open-label”therapy - Preferred that subject stays in trial even if discontinues the intervention - Data and safety monitoring board (DSMB)-early stopping rules

Front 170

What are the key elements of informed consent?

Back 170

- Participation is voluntary! - Confidentiality of information for study subjects - Full disclosure of risks and benefits - Participants must know what to expect

Front 171

What are the 4 phases of clinical trials of drugs?

Back 171

1) Determine the safety, toxicity, tolerability of the drug

2) Determines biologic activity of the new intervention/efficacy/safety

3) Demonstrate therapeutic benefit - efficacy and effectiveness

4) Long-term surveillance

Front 172

How do you set-up a chart to analyze disease screening?

Back 172

Disease + -

Test +

-

Front 173

What is a diagnostic test?

Back 173

Diagnostic testing confirms disease in those suspected to have it

Front 174

What is a screening test?

Back 174

Screening testing looks for disease regardless of symptoms

Front 175

What is a gold standard test?

Back 175

Screening often followed up with “Gold Standard” test◦more invasive◦more expensive◦more uncomfortable

Front 176

What are the 4 purposes of screening?

Back 176

1) To identifypre-disease condition/riskto prevent disease

2) Toidentify disease for early treatment

3) To identify extent of disease burden

4) Toassess effectivenessof prevention programs –especially on community level

Front 177

What is mass screening?

Back 177

Assumes everyone is at risk

Front 178

What is selective screening?

Back 178

Assumes knowledge of who is at risk

Front 179

What are the three types of prevention?

Back 179

1) Primary

2) Secondary

3) Tertiary

Front 180

What is the purpose of screening?

Back 180

"Early detection of a disease or condition in the preclinical phase”

Front 181

When is a disease appropriate for screening?

Back 181

- Disease is an important public health problem◦magnitude◦seriousness◦ Disease has a detectable pre clinical phase◦ Disease must be treatable (or prevention must be key)◦early treatment is better than late treatment

Front 182

What are the Characteristics of a Good Screening Test?

Back 182

- Easy to administer

- Acceptable to patients

- Relatively inexpensive

- Safe - Follow-up facilities available

Front 183

What are the Characteristics of a Good Screening Test?

Back 183

Reliable: Reliability is another term for consistency.

Valid: A test is valid if it measures what it is supposed to measure.

Reliability and validity are independent of each other. A measurement maybe valid but not reliable, or reliable but not valid. Suppose your bathroomscale was reset to read 10 pound lighter. The weight it reads will be reliable(the same every time you step on it) but will not be valid, since it is not reading your actual weight.

Front 184

What is inter-and intra-observer variation?

Back 184

same test results read by different people

Front 185

What are the two ways to measure validity?

Back 185

sensitivity

specificity

Front 186

What is Intra-observer variation?

Back 186

variability in results from same observer at different times

Front 187

Predictive Value

Back 187

◦What is the probability that a test result (positive/negative) is correct?

Front 188

What is sensitivity?

Back 188

Proportion of true positives out of all with disease (true positive rate)

Front 189

What is specificity?

Back 189

Proportion of true negatives out of all without disease (true negative rate)

Front 190

Do Sensitivity and Specificity vary across populations?

Back 190

No

Front 191

What’s wrong with too sensitive?

Back 191

Wasted resources

Health consequences

-from tests

- from treatments

- from emotional cost

Loss of confidence in tests

Front 192

Is it possible for a test to be both sensitive and specific?

Back 192

-yes

Expectedof gold standards and generally requiresmore complicated or even invasive tests

But often a trade off in screening tests

Front 193

What is SNOUT?

Back 193

Very sensitive test+result is not very helpful, but result is useful = rule disease out

sensitivity rules disease out

Front 194

What is SPIN?

Back 194

Very specific test - result is not very helpful, but + result is useful = rule disease in

specificity rules disease in

Front 195

Positive Predictive Value

Back 195

Proportionof true positives out of all who test positive

Front 196

Negative Predictive Value

Back 196

Proportion of true negatives out of all who test negative

Front 197

Potential screening biases?

Back 197

Referral bias

Lead time bias

Length bias sampling

Over-diagnosis bias

Front 198

What is Referral or Volunteer Bias?

Back 198

Can we assume those screened were the same as those not screened?

Observational study Volunteers

◦Maybe healthier = lower mortality

◦Maybe higher risk = higher mortality

Randomized trial can help to minimize this but still have to pay attention to participants

Front 199

What is LeadTime Bias?

Back 199

If there is no true benefit from early detection of disease, there appears to be a benefit from screening because of an earlier point of diagnosis from which survival is measured

Front 200

What is Length Bias Sampling?

Back 200

Screeningtest preferentially identifies those with longer preclinical phase of disease

Comparing outcomes identified by screening vs. usual means

◦Better outcomes in screened group

Front 201

What is the Overdiagnosis Bias?

Back 201

-New screening program = enthusiasm

May result in over-reading of test

Normal people included in diseased group (FP)

Occurs if no blinding

◦Inflated survival after screening because normals included in diseased group

Front 202

Interpret: Sensitivity: 20/30=67%

Back 202

Of those with the disease, 67% had a positive test result. This means that 33% of people with the disease tested negative (33% false negative rate). The test misses 1/3 of those with disease.

Front 203

Intepret: Specificity: 1820/2000=91%

Back 203

Of those without the disease, 91% had a negative test result. This means that 9% of people who were disease free tested positive(9% false positive rate). The test has a very low number of false positives.

Front 204

Interpret: PV+: 20/200=10%

Back 204

If a patient had a positive test, there was only a 10% chance that (s)he actually has colorectal cancer. 90% of the people who test positive won’t actually have the disease.◦If you have a positive test, it’s not very likely that you have colorectal cancer.

Front 205

Interpret: PV-: 1820/1830=99.5%

Back 205

Negative predictive value: If a patient has a negative test, there is a 99.5% chance that (s)he does not have colorectal cancer. Only 0.5% of people who test negative actually have the disease.◦If you have a negative test, it’s highly likely that you do not have colorectal cancer.

Front 206

What is the CDC definition of surveillance?

Back 206

Ongoing, systematic collection, analysis, and interpretation of health-related data essential to the planning, implementation, and evaluation of public health practice, closely integrated with the timely dissemination of these data to those responsible for prevention and control.

Front 207

What is the purpose of surveillance?

Back 207

•Assess public health status.

•Define public health priorities

•Evaluate programs

•Stimulate research

Front 208

What is the purpose of surveillance?

Back 208

•To assess public health status, to define public health priorities, to evaluate programs, and to stimulate research.

◦Tells us where the problems are, who is affected, and where the programmatic and prevention activities should be directed.

◦Legal authority for surveillance resides with states

Front 209

What are the three types of surveillance?

Back 209

- 1) Active: the local or county health department call asking for data - resource intensive - timely - ex) looking at food borne illness through FoodNet.

- 2) Passive: The lab doesn’t call to ask, they hope providers report - ex) a list of notifiable diseases

- 3) Sentinel: We may select certain labs or providers - we are reaching out in a targeted way - ex) GISP is looking for a particular strain of the thing

Front 210

How can surveillance data be used?

Back 210

•Hypothesis testing

•Evaluating control and prevention measures

•Monitoring change

•Detecting changes in health practice.

•Facilitate planning

Front 211

Uses of Surveillance Data:Facilitate Planning

Back 211

•Identify target populations in need of health services

◦Refugee populations

◦Morbidity surveillance in emergency shelters

•Identify health topics to be addressed by educational programs and media

Front 212

What are the three ways surveillance is outcome oriented?

Back 212

•Can measure frequency of an illness or injury (e.g., number of cases, incidence, prevalence)

•Can measure severity of the condition (e.g., hospitalization rate, disability, case fatality)

•Can measure impact of the condition (e.g., cost)

•Orient data by person, place, and time

Frequency, severity, impact

Front 213

How would you plan a Surveillance System?

Back 213

•Establish objectives.

•Develop case definitions

•Determine data source or data collection mechanism

•Field test methods

•Develop and test analytic approach

•Develop dissemination mechanism

•Assure use of analysis and interpretation

Front 214

What Should be Under Surveillance?

Back 214

Establish priorities based on:

◦Frequency (incidence, prevalence, mortality)

◦Severity (case-fatality, hospitalization rate, disability rate, years of potential life lost).

◦Cost (direct and indirect)

◦Preventability

◦Communicability

◦Public interest

◦Will the data be useful for public health action?

Front 215

Surveillance Methods: Why is a Case Definition Important?

Back 215

•Important to clearly define condition.

•Ensures same criteria are used by all

•Makes the data more comparable

•Include person, place, time

•May define suspected and confirmed cases

•May include symptoms, lab values, time period, population as appropriate

Front 216

Three Keys to a Case Definition Example

Back 216

1) clinical description

2) lab criteria for diagnosis

3) case classification

Front 217

Passive surveillance*

Back 217

◦Providers are responsible for reporting

◦Health dept. waits to receive reports

◦Problem with under reporting

Front 218

Active surveillance*

Back 218

◦Providers contacted on regular basis to collect information

◦More resource intensive.

◦Used for outbreaks or pilot studies

Front 219

Surveillance Methods:Data Analysis

Back 219

•Ongoing review

•Descriptive statistics, multivariate analyses. •Automated analyses

Front 220

Data Sources in Surveillance

Back 220

•Vital Statistics.

•Notifiable Diseases.

•Registries.

•Sentinel Surveillance.

•Syndromic Surveillance.

•Surveys.

•Administrative Data

Front 221

What does the Quality of Vital Stats Depends on

Back 221

•Care taken by health care providers in ascertaining cause of death and other factors. •Accuracy of coding (difficult for injuries). •Relevance of existing codes for the condition being recorded

•Accuracy of population estimates

•Problems - don’t know onset, can’t see effect of diseases that don’t lead to death

Front 222

National Electronic Disease Surveillance System (NEDSS)

Back 222

National Electronic Disease Surveillance System (NEDSS)–A set of criteria developed by CDC that all public health surveillance systems must meet

Front 223

Limitations of Disease Reporting

Back 223

Under reporting

◦Reporting better for more serious diseases and those for which there is laboratory confirmation

◦Need to seek medical consultation to be diagnosed and then reported

•Lack of representativeness of reported cases

•Inconsistent case definitions

Front 224

Reasons for Not Reporting

Back 224

•Assume someone else reported.

•Do not know reporting was required; don’t have a copy of the reportable disease list.

•Do not know how to report; don’t have form or telephone number.

•Concern about confidentiality and doctor-patient relationship.

•No incentive to report. Time-consuming. Unaware of value.

Front 225

How to Improve Reporting

Back 225

Contact physicians in the community.

◦Tell them the health department is very interested in morbidity reporting

•Maintain a reasonable list of reportable diseases.

•Maximize contact through presentations, mailings, newsletters, media, etc.

•Use the data.

Front 226

What are Sentinel Systems?

Back 226

- Sentinel means to keep watch

- To gather timely public health information in a relatively inexpensive manner.

- Cannot derive precise estimates of prevalence or incidence in the population.

- Sentinel health events

- Sentinel sites

- Sentinel providers

Front 227

What is a Sentinel Health Event?

Back 227

- A condition whose occurrence serves as a warning signal.

- Particularly useful for occupational exposures.

- Examples: Silicosis, pesticide poisoning, lead poisoning, occupational asthma

- Cases trigger intervention activities.

Front 228

What are Sentinel Sites or Providers?

Back 228

- Surveillance at certain hospitals, clinics, or physician practices.

- Sentinel sites

- monitor conditions in subgroups that may be more vulnerable

◦E.g.,drug clinic, STD clinic, MCH clinic

-Sentinel providers - monitor activity in ambulatory care settings.

◦For diseases that are not reportable

Front 229

What is Syndromic Surveillance?

Back 229

“The collection and analysis of health-related data that precede diagnosis and signal a sufficient probability of case or an outbreak to warrant further public health response.”

Uses pre-diagnostic indicators to identify emerging health problems

Front 230

What are syndrome categories?

Back 230

–Death

–Sepsis(serious infection)

–Rash–Respiratory(e.g., cough)

–Gastrointestinal(e.g., diarrhea)

–Unspecified Infection (fever)

–Neurological(e.g., dizziness)

Front 231

What are uses of Syndromic Data?

Back 231

- Monitor trends in influenza, gastrointestinal illness

- Detect outbreaks or individual cases of disease, especially illnesses with unique symptoms or names

- Special event surveillance (e.g., Olympic Games, Presidential Inauguration, National Boy Scout Jamboree)

- Disaster surveillance (e.g., hurricanes, ice storms, etc.)

Front 232

Data Source: Surveys

Back 232

- If done continually or periodically, can monitor risk factors and changes in prevalence over time

- Can also assess knowledge, attitudes

- People usually queried only once and not monitored on an individual basis after that

- From questionnaires, interviews (in person or telephone), or record review

Front 233

Data Source: Administrative Data

Back 233

- Routinely collected for other reasons.

- Hospital discharge data collected for billing purposes, Medicaid and Medicare data,emergency department data, data collected by managed care organizations.

Front 234

What does the usefulness of Administrative Data depend on?

Back 234

- Depends on:

◦What information is computerized

◦Standardization of codes for diagnoses, symptoms, procedures, reasons for the visit

◦Time between occurrence of health event and availability of data

◦Ability to link with other data systems

◦Whether supplementary information can be obtained

Front 235

What is adverse Event Surveillance?

Back 235

- Focuses on patient safety

- Passive surveillance

Front 236

What are interpretations of Surveillance Data?

Back 236

- Limitations

◦Under reporting

◦Biased reporting

◦Inconsistent case definitions

- Consider context

◦Seasonality

◦Recent policy changes or interventions

Front 237

How do you evaluate Surveillance Systems?

Back 237

- System objectives and usefulness

◦Actions taken as a result of the data.

◦Does the system do what it’s supposed to do? - Operation of the system

◦Who is reporting? To whom? What information is collected? How is information stored? Who analyzes the data? What are the findings? How often are reports disseminated? to whom?

- Cost

Front 238

Evaluation - System Attributes

Back 238

- Simplicity

◦Should be as simple as possible and as easy to operate as possible.

- Flexibility

◦Should be able to adapt to changing needs.

- Acceptability

◦Willingness of individuals or organizations to participate in the surveillance system. (Judge based on completeness, timeliness, reporting)

- Sensitivity

◦Proportion of cases detected by the system. Completeness of reporting. Detect epidemics?◦Increased awareness, new diagnostic test, change in surveillance method may impact. - Predictive Value Positive

◦Proportion of persons identified as having the disease who actually have it.

- Representativeness ◦Do the characteristics of reported events compare favorably with those in the population.

◦Is there case ascertainment bias?

◦Bias in descriptive information about a reported case? - Timeliness

◦Any delay between the steps? (onset, diagnosis, report to public health, disease control actions)

Front 239

Process vs. Outcome Measures

Back 239

- Process measures - determine components of good care and determine extent to which program meets criteria

◦Do not indicate quality of care or whether person benefits

◦ Criteria for what is “good” may be variable ◦ Outcome measures

- determine whether person benefits

◦Clearly quantifiable

◦Easy to define

◦Standardized measure

Front 240

Screening Program Effectiveness Process Measures

Back 240

- Number screened - Proportion of target population screened

- Total cost

- Cost per case found

- Proportion of positive screens diagnosed and treated

- Sensitivity and specificity and positive predictive value season

Front 241

Screening Program Effectiveness Outcome Measures

Back 241

- Reduction in mortality in screened population

- Reduction in case fatality in screened individuals - Increase in early stage cases - Reduction in complications - Prevention of recurrence or disease progression

- Improvement in quality of life in screened individuals

Front 242

What is a confounder and why does it matter?

Back 242

Occurs when the observed result between exposure and disease differs from the truth because of the influence of the third variable

In epidemiologic terms, the tobacco companies were claiming that air pollution (or any other factor that can cause cancer) is a confounding variable. A confounding variable is a variable (say, pollution) that can cause the disease under study (cancer) and is also associated with the exposure of interest (smoking).

Confoundingneeds to be considered as a potential explanation of findings from ~anyobservational study!

Front 243

Confounding “General Rules”

Back 243

1) Associatedwith the exposure (but not a consequence of the exposure)

2) Associatedwith the outcome, causally or non-causally (e.g., arisk factor)

3) Not in the causal pathway between the exposure and outcome

Front 244

Is the confounder associated with the exposure of interest what do you do?

Back 244

Calculate “informal” OR to assess association

Front 245

How do you quantitatively compare stratum for confounders?

Back 245

1) Compare stratum-specific OR/RR to each other and to the crude OR/RR

- If stratum specific OR/RR are similar to each other and different from crude CONFOUNDING- Similar”can be interpreted broadly! Should be in the same direction, away from the crude.

- Ifstratum specific OR/RR are similar to each other and similar to the crude

3) NOCONFOUNDING

4) Compareadjusted OR/RR to crude OR/RR

If different CONFOUNDING

Front 246

Whatare we going to do about confounding?

Back 246

- Design

◦Randomization

◦Restriction

◦Matching

- Analysis

◦Stratification

◦Adjustment

Front 247

How do you use matching to Minimize Confounding?

Back 247

Do a case control study

Front 248

How do you restrict to Minimize Confounding?

Back 248

Cohort Study - Women Only

Front 249

How do you calculate the odds Ratio in a Matched Case-Control Study?

Back 249

Paired Odds Ratio = b: number of case: control pairs where case has exposure, control is unexposed/c: number of case: control pairs where control has exposure, case is unexposed

Front 250

How do you set-up the table in an oddsRatio: Matched Case-Control Study?

Back 250

Control is Exposure and Control is Unexposed vs. Case is exposed vs. Case is unexposed [In a chart]

Front 251

How do you control in a confounding analysis?

Back 251

- Stratification - Adjustment

◦Rate standardization

◦Mantel-Haenszel adjusted odds ratio

◦Multi-variable regression models

- Interpret all findings with caution

Front 252

Definition of Interaction/ Effect Modification

Back 252

Does the strength of our association of interest ~differ~ in the presence of a third factor?

Front 253

What is Multiplicative vs. Additive Interaction?

Back 253

- Additive interaction

◦ Assumes two exposures each increase risk independently

- Multiplicative interaction

◦Assumes two exposures increase risk synergistically

◦“The whole is greater than the sum of its parts.”

- Dependent on biological mechanisms

- We will focus on multiplicative interaction

◦Much more common in the literature

◦Facilitated by our statistical tools

Front 254

Rules about interaction vs. effect modification vs. confounding

Back 254

- When confounding is present,stratum

-specific RR/OR will be similar to each other - Confounders make it difficult to assess causality

- Confounding is a nuisance that we try to minimize

- Confounding should be considered before, during, and after analysis

- There is no statistical test to assess for confounding

Front 255

What is the difference between confounding and interaction

Back 255

- Confounding

◦Nuisance to eliminate

◦Not real

- Interaction

◦Natural phenomenon to describe and understand

Front 256

Method 1 for Confounding Review - Questions

Back 256

◦Is confounder associated with exposure? ◦Is confounder associated with outcome? ◦Is confounder on the causal pathway? ◦Must be associated with exposure and outcome and not on the causal pathway in order to be a confounder

Front 257

Method 2 for Confounding Review - Crude vs. Strata Specific

Back 257

◦Stratify by the confounder and calculate measures of association

◦Compare the crude measure of association to strata specific

◦If crude is different from strata specific, then variable is confounder.

◦Note strata must be similar to each other.

Front 258

Method 3 for Confounding Review - Crude vs. Adjusted

Back 258

◦Compare the crude measure of association to adjusted measure

◦If crude is different from adjusted, then variable is confounder.

◦Note strata must have ruled out effect modification.

Front 259

Ways to control confounding in design and analysis

Back 259

- Design

◦Randomization

◦Restriction

◦Matching

- Analysis

◦Stratification

◦Adjustment

Front 260

What is Bias?

Back 260

“Any systematic error in the design,conduct or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of disease.” (Gordis)

Front 261

Bias vs. Confounding vs. Interaction

Back 261

- BIAS:Incorrect association due to a flawed sample selection or data collection method - CONFOUNDING:Incorrect association due to a third variable’s association with both the risk factor and outcome (but not a causal factor)

- INTERACTION:Effect is real but the magnitude of the effect is different for subgroups (e.g.gender, race) s Unicode"

Front 262

What are sources of bias?

Back 262

◦Can arise in all study types

•◦Occurs in design and conduct of a study

•◦Can be evaluated but not fixed in the analysis phase

•◦Two main types of bias are selection and information bias

Front 263

What are effects of bias?

Back 263

- Bias can create spurious associations when there really is none

◦Bias away from the null

- Bias can mask an association when there really is one

◦Bias towards the null

Front 264

What is Selection Bias?

Back 264

Results when the procedures used to choose study subjects lead to a result different from what would have been obtained if the entire population had been included in the study.

Front 265

Selection Bias in Case Control vs. Cohort Study

Back 265

- Case-Control:

◦Factor studied: Exposure

◦Selection based on: Disease

◦OR will be biased if selection differs between E/Non E cases or between E/Non E controls

- Cohort:

◦Factor studied: Disease

◦Selection based on: Exposure

◦RR will be biased if selection differs between D/Non D among exposed or between D/Non D among non-exposed

Front 266

Selection Bias in a Cohort Study

Back 266

- Selection bias occurs when selection of exposed and unexposed subjects is not independent of outcome

◦Differential completeness of records (retrospective cohort)

◦Differential loss to follow-up (prospective cohort)

◦Differential ability to measure outcome of interest (prospective cohort)

◦Differential participation (prospective cohort)

Front 267

What are Special Cases of Selection Bias?

Back 267

- Healthy Worker Effect

- Berkson bias (hospital controls)

- Neyman bias (incidence-prevalence bias)

- Unmasking bias (detection signal)

Front 268

Healthy Worker Effect

Back 268

- Generally,working individuals are healthier than individuals who are not working - In occupational exposure studies, where cases are workers, controls should also be workers ◦Otherwise,association between exposure-disease will be biased towards the null Same is true for exposed and non-exposed workers

Front 269

Berkson’s Bias

Back 269

Case-control study of smoking and lung cancer - Controls are selected from a hospital population

◦Association between exposure and disease weakens – WHY?? Smoking causes many diseases resulting in hospitalization Hospital controls do not represent the prevalence of exposure (smoking) in the community from which cases of lung disease arise Better to sample controls from the community.

Front 270

Neyman Bias (Incidence-Prevalence)

Back 270

Ex) snow shoveling and heart attack - hospital based study, would miss individuals who never reached the hospital, therefore the association between the exposure and the disease is underestimated

Front 271

Unmasking Bias

Back 271

Case-control study of oral contraceptive use and diabetes - diabetics visit the hospital more, therefore, exposed cases have a higher probability of being included in the study

Front 272

Minimizing Selection Bias - bad news/good news

Back 272

The bad news:

•Selection bias cannot be fixed once it has occurred

•The good news: You can design and conduct your study in ways that will minimize the likelihood of selection bias

−Use similar criteria for selecting cases and controls (and exposed and unexposed) −For hospital-based controls: consider diagnostic and referral patterns

−Design data collection systems to minimize missing data

−Develop protocols to ensure high participation/follow-up rates

Front 273

Selection Bias vs. Selecting Study Subjects

Back 273

- Selection Bias

◦Did our choice and selection of comparison groups result in a valid estimate? ◦Threat to internal validity

◦ Selecting Study Subjects / Sampling Error ◦Are the findings from our sample generalizable to the population?

◦Potentially affects external validity

Front 274

When does information bias occur?

Back 274

- An error that occurs when the methods for obtaining data about study subjects are inadequate and some participants are incorrectly classified on their:

◦Exposure status

◦Outcome status

Front 275

Non-Differential Misclassification

Back 275

- Exposure is misclassified similarly between outcome groups

- Outcome is misclassified similarly between exposure groups

- Either EVERYONE in the study is affected, or a similar proportion in each comparison group is affected

- Non-differential misclassification weakens the exposure-outcome association

◦Result:Will bias estimate (OR/RR) toward the null

Front 276

Differential Misclassification

Back 276

- Occurs if the degree of misclassification differs between comparison groups ◦Particularly problematic if data collection procedures are different between groups

- The effect of differential misclassification is unpredictable and depends on the specific research question

◦Result: May bias the association(RR/OR) either away from or towards the null

Front 277

Special Cases of Information Bias

Back 277

- Recall bias

- Interviewer bias

- Differential loss to follow-up

Front 278

Recall Bias

Back 278

- A specific type of differential misclassification in case-control studies

- People with disease remember or report exposures differently (more or less accurately) than those without disease

- Can result in over- or under estimate of the measure of association

- Different from (overall) recall error!

Front 279

Solutions for reducing recall bias

Back 279

◦Use objective measures whenever possible

◦Use standardized questionnaires that obtain detailed information

◦Use controls who have a different outcome (also ill)

◦Blind subjects to study hypothesis

Front 280

Interviewer Bias

Back 280

- Systematic difference in soliciting, recording, interpreting information

- Occurs when exposure information is sought when the outcome is known (case-control) or when outcome information is sought when exposure is known (cohort)

Front 281

Minimizing Information Bias

Back 281

- Use objective, calibrated, validated, standardized assessment tools

- Maximize follow-up in all comparison groups - Minimize missing data in all comparison groups

- Blind study staff to exposure and/or disease status

- Mask participants to study hypothesis

- Some misclassification is (usually) unavoidable. Non-differential misclassification is preferable to differential misclassification!

Front 282

Other Types of Bias

Back 282

◦Hawthorne effect

◦Surveillance bias

◦Publication bias

Front 283

Bias Summary

Back 283

When interpreting study results, ask yourself these questions:

1.What are the potential sources of bias in this study?

a.Selection bias – Are my comparison groups from the same underlying population? Were there systematic differences in recruitment/follow-up/participation between groups?

b.Information bias – Is my exposure measured correctly/completely? Is my outcome measured correctly/completely? Is measurement similar between comparison groups?

2. What is the potential magnitude of the bias?3. Which direction is the distortion? Is it towards the null or away from the null?

Front 284

Why Compare Risk Between Groups?

Back 284

- Measures of excess risk between groups

◦Relative differences (RR, OR)

◦Absolute differences (AR, AR%, PAR, PAR%)

- Identify disease etiology-relative differences (RR)

- Estimate public health impact of exposure prevention-absolute differences (AR)

Front 285

Attributable Risk / Population Attributable Risk

Back 285

- Includes: AR, AR%, PAR, PAR%

◦Amount or proportion of disease incidence attributed to exposure

◦How much disease can we hope to prevent if we eliminate the exposure?

- Estimates the absolute excess risk associated with a given exposure

◦Implies a cause-effect association

◦This is a big assumption!

- Important in the context of public health decision-making and resource allocation

Front 286

Attributable Risk (AR) among Exposed

Back 286

- Excess risk of disease due to the exposure among those who are exposed “Among smokers, what incidence of lung cancer is due to smoking?”

Front 287

Attributable Risk (AR) among Exposed

Back 287

- If all smokers stopped smoking,we estimate a reduction of lung cancer rates of 8/100,000 among smokers fall smokers quit, we could potentially prevent 8/100,000 cases of lung cancer among smokers.

Front 288

AR Percent (AR%) among Exposed

Back 288

- Percentageof excess risk of disease due to the exposure amongthose who are exposed “Among smokers, what proportion of lung cancer incidence is due to smoking?”

A few extra notes…

- AR can only be directly calculated when the study design provides incidence

- AR%is equivalent to: (RR-1) / RR

◦Can be extended to case-control studies (OR≈RR)

Front 289

Population Attributable Risk (PAR)

Back 289

- Excessrisk of disease due to the exposure among the entire population “What incidence of lung cancer is due to smoking?”

◦If all smokers quit, we anticipate a reduction in lung cancer rates of 2/100,000 in the population. We could potentially prevent 2 of 4/100,000 incident cases of lung cancer in total population if all smokers quit.

Front 290

PAR Percent (PAR%)

Back 290

- Percentageof excess risk of disease due to the exposure amongthe entire population “What proportion of lung cancer incidence is due to smoking?”

A few extra notes…

- Like AR, PAR can only be directly calculated when the study design provides incidence

◦Can be extended to case-control studies (OR≈RR)

Front 291

Absolute vs. Relative Risks: Summary

Back 291

- RR/OR/HR: Doesexposure (E) cause disease (D)?

- AR: Amongpersons exposed to E, what amount of the incidence of D is E responsible for?

- AR%: Among persons exposed to E, what proportion of the occurrence of D was due to E?

- PAR: Whatincidence of D in the population is due to E? (Shouldresources be allocated to controlling E or, instead, to exposures causinggreater health problems in the population?)

- PAR%: What portion of D in the population is caused by E? (Should resources allocated to combating D be directed toward etiologic research or control of known etiologies (e.g., E)?)

Front 292

Causality

Back 292

- Epidemiology= study of disease etiology (cause)

- Cause (noun): “an antecedent event, condition, or characteristic that was necessary for the occurrence of the disease at the moment it occurred, given that other conditions are fixed”

Front 293

Characteristics of Cause

Back 293

- Must precede the effect

- Can be either host or environmental factors (e.g. characteristics, conditions, actions of individuals, events, natural, social or economic phenomena)

- Positive(presence of a causative exposure) or negative (lack of a preventive exposure)}Not all associations are causal! Being born in Northern Europe is not a cause for breast cancer; marker for populations that may have a greater genetic predisposition to breast cancer

Front 294

What are the limitations for determining a causal association at each step?

Back 294

Clinical Observations - available data - case-control studies - cohort studies - randomized trials

Front 295

Direct vs. Indirect

Back 295

Direct:factor directly causes a disease without any intermediate step •Indirect:factor causes a disease, but only through an intermediate step(s)

Front 296

Causality Vocabulary: Sufficient Cause

Back 296

A set of conditions without any one of which the disease would not have occurred;cause will result in event without any additional event or condition (entire pie)

Front 297

Causality Vocabulary: Component Cause

Back 297

Any one of the set of conditions which are necessary for the completion of a sufficient cause (piece of the pie)

Front 298

Causality Vocabulary: Necessary Cause

Back 298

A component cause that is a member of every sufficient cause; event cannot happen without cause

Front 299

Causality Vocabulary: Sole

Back 299

- Cause is necessary and sufficient for the event ◦One-to-one exposure to disease relationship rarely if ever occurs

◦Without Factor A, disease will never develop (necessary)

◦In the presence of Factor A, the disease will always develop (sufficient)

Front 300

Causality Vocabulary: Direct/Proximal

Back 300

Cause is sufficient and the last step in a causal chain

Front 301

Causality Vocabulary: Indirect/Distal

Back 301

Cause alone is not sufficient, but is in a causal chain that results in a causal sequence that is sufficient

Front 302

Necessary but not sufficient

Back 302

When I was a kid, I remember reading once about a scientist saying that the problem with locating brain functions by what's impaired when somebody has brain damage there, is that it's like opening up a TV set and taking out a resistor. If the picture goes bad, you might then conclude that the resistor is the "source of pictureness", when all you have really proved is that the resistor (or brain part) is necessary for pictureness. Not that it's sufficient.

Front 303

Sufficient but not necessary

Back 303

For example, it is not necessary to earn 950 points to earn an A in this course. You can earn 920 points to earn an A.

Front 304

Rothman’s Sufficient-Component Cause Model

Back 304

- Conceptual model of causation (1976) - General model for the conditions necessary to cause (and prevent) disease - Sufficient cause “a complete causal mechanism” that“inevitably produces disease” - Recognized that disease outcomes have multiple determinants◦ I.e.one set of determinants that produce TB in one individual may not be the same set of conditions that were responsible for the occurrence in others

Front 305

Attributes of the Causal Pie

Back 305

- Completion of a sufficient cause is synonymous with occurrence (not always diagnosis) of disease

- Component causes can act far apart in time - A component cause can involve the presence of a causative exposure or lack of a preventive exposure

- Blocking the action of any component cause prevents the completion of the sufficient cause and therefore prevents the disease by that pathway

Front 306

Hill’s Guidelines

Back 306

- Published in the first Surgeon General Report on Smoking and Health

◦Was the statistical association between smoking (exposure) and lung cancer(outcome) causal?

◦Highlights minimal conditions to establish causality

◦No single criterion -- or even all criteria -- is sufficient to determine causality

◦Inspires discussion, debate, and modifications!

Front 307

Hill’s Guidelines - 9 things

Back 307

1.Temporal relationship

2.Strength of the association

3.Dose-response relationship

4.Replication of findings

5.Biologic plausibility

6.Consideration of alternate explanations

7.Cessation of exposure (experimental)

8.Consistency with other knowledge

9.Specificity of the association

Front 308

Temporal Relationship

Back 308

- The causal factor must precede the disease in time

- This is the only one of Hill’s criteria that everyone agrees with

- Prospective studies do a good job establishing the correct temporal relationship between exposure/disease

◦ Example:prospective cohort study of smokers and nonsmokers, follows them to determine the occurrence of subsequent lung cancer

Front 309

Strength of Association

Back 309

- The larger the association, the more likely the exposure is causing the disease

◦RR of lung cancer in smokers vs. nonsmokers = 9

◦RR of lung cancer in heavy smokers vs. nonsmokers = 20

- Strong associations are more likely to be causal because they are unlikely due entirely to bias and confounding

- Weak associations may be causal but it is harder to rule out bias and confounding

Front 310

Dose Response Relationship

Back 310

- Persons who have increasingly higher exposure levels have increasingly higher risks of disease

◦Lung cancer death rates rise with number of cigarettes smoked

•Some exposures might not have a dose-response effect but rather a “threshold effect” below which these are no adverse outcomes

- When present,dose-response provides evidence of causality, but in its absence does not preclude causation

Front 311

Replication of Findings

Back 311

- The association is observed repeatedly in different persons, places, times, and circumstances

- Replicating the association in different samples,with different study designs, and different investigators gives evidence of causation

◦Example: Smoking has been associated with lung cancer in at least 29 retrospective and 7 prospective studies

Front 312

Biological Plausibility

Back 312

- Biological or social model exists to explain the association

- Association does not conflict with current knowledge of natural history and biology of disease

◦Cigarettes contain many carcinogenic substances

Front 313

Alternative Explanations

Back 313

- Have all other possible explanations been taken into account?

- Have all other possible explanations been ruled out?

Front 314

Cessation of Exposure

Back 314

- Does the risk of disease decline when exposure to the factor is reduced or eliminated? - Investigator-initiated intervention that modifies the exposure through prevention, treatment, or removal should result in less disease

◦Smoking cessation programs result in lower lung cancer rates

Front 315

Consistency with Other Knowledge

Back 315

- Are the findings consistent with other data? ◦Consistent direction in curves, with increasing rates of lung cancer following increasing cigarette sales for both men and women

Front 316

Specificity

Back 316

- A single exposure should cause a single disease

- From concepts of causation that were developed for infectious disease (many exceptions to this), weakest criteria ◦Smoking is associated with lung cancer as well as many other diseases. In addition, lung cancer results from smoking as well as other exposures.

- When present,specificity provides evidence of causality, but in its absence does not preclude causation