150A Midterm studyguide

Is a Disease Associated with an Exposure?

Must determine whether there is a greater risk of the disease in persons who have been exposed to certain agent than persons not exposed

Data from case-control and cohort study designs most frequently used

Both case-control and cohort studies are designed to determine whether there is an association between exposure and disease by making comparisons; and assessing the magnitude of the association

Estimating Risk

Defined as the probability of an event occurring (such as development of disease or death over a specified observation window)

Objective of most study designs in epidemiology is to determine whether there is an increased risk (e.g., incidence) or decreased risk of a certain disease in association with a certain exposure or other characteristic

I am going to focus on incidence for the examples here— prevalence is thorny. Researchers are usually interested in things that cause onset or acquisition of disease. Prevalence is affected by this, but can also be affected by disease duration.

Absolute Risk

Incidence of disease (mortality)in a population

Can indicate magnitude of risk in a group of people with a certain exposure BUT does not take into account the risk of disease in non‐exposed individuals

Does not stipulate any explicit comparison

Comparison is fundamental to epidemiology

Measures of Association

Calculated in two ways

Difference in absolute risks (incidence or death) = absolute measure

Subtraction

Ratio of absolute risks (incidence or death) = relative measure

Division

Summary of Measures of Association: Relative Risk

Incidence density ratio

cumulative incidence ratio

odds ratio

Incidence density ratio (IDR)

Relative rate; Rate ratio = ID exposed/ID nonexposed

Interpretation: The rate of {outcome} in the {exposed} is {#} times the rate in the {unexposed}.

Cumulative incidence ratio (CIR)

Relative Risk; Risk ratio = CI exposed/CI nonexposed

Interpretation: The risk of {outcome} in the {exposed} is the {#} times the risk in the {unexposed}.

odds ratio (OR)

Odds exposure diseased/ Odds exposure not diseased

AD/BC

If OR = 1

• Odds of exposure in cases = odds in control (no association)

If OR > 1

• Odds of exposure in cases > odds of exposure in controls (positive association)

If OR < 1

• Odds of exposure in controls > odds of exposure in cases (negative association)

types of relative risks

Cumulative Incidence Ratio (CIR or Risk Ratio)

Incidence Density Rate Ratio (IDR or Rate Ratio

Cohort study

Our Friend the 2x2 Table

Important Measures for Evaluating Disease Impact

Absolute risk

Absolute risk (probability of disease or death in total population, or Cumulative Incidence in total population)

incidence of disease in the total population

A + C / (TOTAL)

does not indicate if exposure is associated with risk of disease (or death)

Estimating Relative Risk - Ratio of Risks

Incidence of disease among exposed persons

a /(a + b) where "a" represents cases of disease in exposed persons, a + b

Estimating Relative Risk - Ratio of Risks

Incidence of disease among unexposed persons

c /(c + d) where "c" represents cases of disease in unexposed persons, c + d

Relative risk ratio

incidence in exposed /incidence in unexposed

Is There an Association? RR

In the hypothetical example, RR=1.61, thus the risk of cancer among cellphone users is 1.61 times the risk of cancer among non‐cellphone users. Another interpretation is that the risk of cancer among cellphone users is 61% higher the risk of cancer among non‐cellphone users

Estimating Relative Risk: Case‐Control Study

Cannot estimate risk of disease directly from case‐control study

To directly estimate disease risks must have values for incidence of disease in exposed and incidence in the non‐exposed

In a case‐control study, we start with disease (cases) and non‐diseased (controls) individuals

Question asked in Case‐control Study: "What are the odds that a case was exposed?"

Understanding Odds

Odds = number of times event occurred /number of times event did not occur

probability of event occurring (p) /probability of event not occurring (1-p)

Odds vary from 0 [nobody has the event—zero in the numerator] ...

...to infinity [everyone has the event—zero in the denominator]

Estimating Risk from a Case‐control Study

Can calculate an exposure odds ratio

odds of exposure in diseased individuals/ odds of exposure in control individuals

= (a/c) / (b/d) = (ad) / (bc)

Odds (disease|exposure) (exposure|disease)

a/b and c/d

a/c and b/d

OR Interpretation

In the hypothetical example, OR=1.62, thus the odds of cellphone use among those with cancer are 1.62 times the odds of cellphone use among those without cancer. Another interpretation is that the odds of cellphone use among those with cancer are 62% higher than the odds of cellphone use among those without cancer.

If OR = 1

• There are the same number of cases that are exposed as controls that are exposed.

If OR>1

• There are more pairs in which the case were exposed, and the controls were not relative to pairs in which the control was exposed, but the controls was not (cases had exposure more often)

If OR<1

• There are more pairs in which the controls was exposed and the case was not relative to pairs in which the case was exposed by the control was not (control exposed more often)

Calculating Odds Ratios in Matched-Pairs Case-Control Studies

Controls are selected by matching each one to a case according to variables that are known to by related to disease risk, i.e. age, sex, race/ethnicity

Results are then analyzed in terms of case‐control pairs rather than for individual subjects

With exposures classified as dichotomous (exposed versus non‐exposed), four types of case ‐control pairs are observed

Pairs may be concordant or discordant with regard to exposure

Types of Pairs in Matched Case‐ control Study

Concordant Pairs

1. Pairs in both the cases and the controls were exposed

2. Paris in which neither the case nor the control was exposed

Types of Pairs in Matched Case‐ control Study

Discordant Pairs

1. Pairs in which the case was exposed but the control was not

2. Pairs in which the control was exposed and the case was not

Summary 2x2 Table: Matched Case‐ control study

Matched pairs odds ratio = b/c

Depression

1 in 6 adults will have depression at some time in their life

Each year ~16 million American adults experiencing depression

More on Symptoms

Persistent sad, anxious, or "empty" mood

Feeling irritable, easily frustrated, or restless

Feelings of guilt, worthlessness, or helplessness

Not wanting to do activities that used to be fun

Decreased energy or fatigue

Moving or talking more slowly

Difficulty concentrating, remembering details, or making decisions

Difficulty sleeping, early‐morning awakening, or oversleeping

Appetite and/or weight changes

Thinking about suicide or hurting yourself

Aches or pains, headaches, cramps, or digestive problems without a clear physical cause and/or that do not ease even with treatment

Diagnosis & Treatment of Depression

In the US, mental disorders are commonly diagnosed using the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), which is published by the American Psychiatric Association.

There are many effective treatments for depression based on the severity. They include:

Psychosocial therapy such as cognitive‐behavioral, interpersonal or problem‐solving

Psychosocial support

Anti‐depressant medication

Depression= major depressive disorder=clinical depression

Common and serious mood disorder

Characterized by experiencing persistent feelings of sadness and hopelessness and losing interest in activities one once enjoyed.

The DSM‐5 outlines the following criterion to make a diagnosis of depression. The individual must be experiencing five or more symptoms during the same 2‐week period.

• Depressed mood most of the day, nearly every day.

• Markedly diminished interest or pleasure in all, or almost all, activities most of the day, nearly every day.

• Significant weight loss when not dieting or weight gain, or decrease or increase in appetite nearly every day.

• A slowing down of thought and a reduction of physical movement (observable by others, not merely subjective feelings of restlessness or being slowed down).

• Fatigue or loss of energy nearly every day.

• Feelings of worthlessness or excessive or inappropriate guilt nearly every day.

• Diminished ability to think or concentrate, or indecisiveness, nearly every day.

• Recurrent thoughts of death, recurrent suicidal ideation without a specific plan, or a suicide attempt or a specific plan for committing suicide.

at least one of the symptoms should be either (1) depressed mood or (2) loss of

How Do We Get Data on Mental Health?

The Substance Abuse and Mental Health Services Administration (SAMHSA) is the agency within the U.S. Department of Health and Human Services that leads public health efforts to advance the behavioral health of the nation. SAMHSA's mission is to reduce the impact of substance abuse and mental illness on America's communities.

National and state‐level data

Tobacco, alcohol, illicit drugs

NSDUH began in 1971, annual

50 states and the District of Columbia

Ages 12 and older

WHO Mental Health Action Plan 2013-2030 (4 objectives)

• Strengthen effective leadership for mental health

• Comprehensive, integrated mental health care

• Implement strategies for promotion and prevention

• Strengthen information systems, evidence, and research

Leaders Care to Prevent Risks

Recurrent depressive disorder

People with depressive episodes experience sadness, reduced energy, feelings of guilt or low self‐ worth for at least 2 weeks

Episodes can be categorized as mild, moderate or severe

Persistent depressive disorder

Depression lasting for at least 2 years (may have periods of less severe symptoms)

Bipolar affective disorder

Consists of both manic and depressive states separated by periods of normal mood

Manic episodes involve elevated mood and over‐activity

Depressive episodes are similar to the unipolar depression

Postpartum depression

Full‐blown major depression during pregnancy or after delivery

Seasonal affective disorder

Onset of depression during the winter months when there is less natural light

Risk Factors for Depression

Major risk factors include: • family history • major life events, trauma or stress • Certain physical illnesses and medications

Measuring Depression

Patient Health Questionnaire is used to monitor the severity of depression (9 Qs)

Center for Epidemiologic Studies Depression Scale (CES‐D) is a screening test used to measure the depressiveness (20 Qs)

Global Trends of Depression

More than 332 million people live with depression around the world (5.7% of adults)

Depression is the leading cause of disability worldwide

It often starts at a young age and affects women more commonly than men

Significant barriers to treatment 1/3 ppl receive mental support

76‐85% of people in low/middle income countries receive no treatment for their disorders

Beyond Treatment: Daily Steps to Manage Feeling Blue Here are other tips that may help you or a loved one during treatment for depression:

Try to be active and exercise • Set realistic goals for yourself • Try to spend time with other people and confide in a trusted friend or relative • Try not to isolate yourself, and let others help you • Expect your mood to improve gradually, not immediately • Postpone important decisions, such as getting married or divorced, or changing jobs until you feel better Discuss decisions with others who know you well and have a more objective view of your situation • Continue to educate yourself about depression

Odds Ratio in Cohort Studies

Concept of odds can be applied to both cohort and case‐control studies

Event happening / Event not happening

In addition to addressing question of "what is the probability (p) that the disease will develop in an exposed person?", i.e., a/a+b, can also ask, "what are the odds that the disease will develop in an exposed person?", i.e,. a:b or a/b, and "..the odds that the disease will develop on an unexposed person", i.e., c:d?

Rare Disease Assumption

If incidence for the disease is low, then:

a+b ~ b

c+d ~ d

Odds Ratio=Relative Risk in Cohort Study: Rare Disease Assumption

A. Calculation of OR in a cohort study B. Calculation of OR in a case‐control study

When is the Odds Ratio a Good Estimate of the Relative Risk for a Case‐control Study?

When three conditions are met:

When cases are representative - with regard to exposure - of all people with the disease in the population from which the cases were drawn.

When controls are representative‐ with regard to exposure‐ of all people w/o the disease in the population from which the cases were drawn.

When the disease is rare (or does not occur frequently)

Summary Of Relative Measures of Association and Study Design

Statistical Significance

For relative measures of association, statistical significance is achieved if:

p‐value <0.05

Confidence interval does not include 1 (for a ratio)

Confidence interval does not include 0 (for difference)

"Statistical significance" signifies that the results would be unlikely (low probability—low p‐value) to have arisen due to chance

Example: Is a Disease Associated with an Exposure?

Association is present when the exposure occurs together with the outcome more often than chance

Association ≠ Causation (may be causal)

Statistical analysis is used to help interpret study results

Measures of Association

Relative risk measures

Strength of association

Disease etiology

Absolute measures of association

Potential impact of eliminating an exposure

Another Measure of Risk: Attributable Risk

Addresses the question, "How much of the disease that occurs can be attributed to a certain exposure?"

Defined: "Amount or proportion of disease incidence (or disease risk) that can be attributed to a specific exposure"

Measure of "impact"

AR implies not all of the disease incidence is due to a single specific exposure, as disease develops in some non‐exposed individuals

Attributable Risk and Public Health Impact

Attributable risk is important in public health and the allocation of resources for improving population health

Fundamental Question: "How much of the risk (incidence) of disease can we hope to prevent if we are able to eliminate the exposure?"

The leading causes of death are a result of a myriad of risk factors. Which factors should we focus on?

Which factors may contribute to the biggest public health impact?

Estimating Potential for Prevention: Examples of AR

1. Can be calculated for exposed persons, e.g., the attributable risk of lung cancer in smokers (attributable risk for exposed persons)

2. Can also be calculated for the total population. e.g., the attributable risk of lung cancer in the total population which includes smokers and non‐smokers (attributable risk for population)

AR We Want to Know:

How much of total risk of disease is actually due to exposure?

Non‐exposed persons also have some risk of disease

Total risk of disease in exposed individuals is the sum of the background risk that any person has and the additional risk due to exposure of interest

Attributable Risk Measures: general formula

Incidence attributable to exposure

Interpretation: Assuming causality, [X] cases of [outcome] among [exposed] is attributable to [exposure].

What proportion of the risk in exposed persons is due to exposure? PAR

((Incidence in exposed group) - (Incidence in non-exposed group)) / Incidence in exposed group

Proportion of incidence attributable to exposure Attributable Risk (Rate) Proportion AR proportion can also be expressed as

What does it mean?:

• 10.6 of the 28/1,000 incident cases in those who smoke are attributable to the fact that they smoke

• Can also be expressed as a proportion or %:

• 37.9% of the morbidity from CHD among those who smoke may be attributable to smoking and could presumably be prevented by eliminating smoking

We must know incidence of the (CHD) disease in the total population OR:

1. The incidence among those who smoke

2. The incidence among non‐smokers

3. The proportion of the total population that smokes

PAR Interpretation

What does this tell us? 21.3% of the incidence of CHD in the total population can be attributed to smoking. An effective program (at best) could achieve reduction of 21.3% in the incidence of CHD in TOTAL population

Interpretation: Assuming causality, [x] cases of [outcome] among the total population is attributable to [exposure].

Measures of Association: Recap of Some Examples

Validity of Study Designs: A Review

Validity of a study: "the degree to which the inference drawn from a study is warranted, when account is taken of the:

study methods - the representativeness of the study samples - and the nature of the population from which it is drawn

Two components: internal and external validity

Internal validity refers to the design, methods, analysis and results of an individual study

Proper selection of study groups and a lack of error in measurement

Study done properly and therefore findings are valid in the study population

Internal validity is a requirement for external validity

External validity refers to the ability to generalize beyond a set of observations to some universal statement

"A study is externally valid or generalizable if it allows unbiased inferences regarding some other target population beyond the subjects in the study"

Internal and External Validity....

Are both basic concerns in the conduct of any experimental trial AND:

Other types of study designs! Like observational studies

Internal Validity: Issues to Consider

Selection of cases or controls or exposed and non-exposed groups

Exposure:

Outcome: disease, behavior, intermediate marker

Association between exposure and disease:

After a MOA is calculated...and an association is observed

The first question to ask is: Is it real?

in order to say anything about: causal inference or external validity ......our study must be internally valid!

We need to rule out:

Bias

Random Error

Confounding

What is Bias?

Systematic error in the design, conduct or analysis of a study that results in an incorrect or invalid estimate of an exposure's effect on the risk of disease....measure of association

Can be introduced at any stage of a study

Can occur in all types of epidemiologic studies

Retrospective studies more susceptible

Two main types: selection bias and information bias

Bias study map

Types of bias

Selection bias: occurs during the selection and follow-up of study participants

Information bias: occurs during data collection

Investigators must:

Identify source of bias - Estimate its magnitude or strength - Assess direction

Selection Bias (systematic error)

Study population does not represent the target population

Results from procedures used to select subjects and from factors that influence participation in the study

Systematic differences in sample

• between cases and controls in a case-control study

• between exposed and non-exposed in a cohort study

• between treatment arms in RCT

Non-compliance example

Selection Bias in Experimental Trial

People who do not comply or do not participate in studies differ from those who do comply and participate

Review example in Table 10.7 and Table 10.8: Coronary Drug Project

Demographic, socioeconomic, cultural, lifestyle medical characteristics

How does Selection Bias Occur?

Control selection bias

inappropriate control group

Non-response or refusal bias

agreement to participate is related to exposure and disease/outcome status (self-selection bias)

Differential surveillance, diagnosis or referral of study subjects

according to exposure and disease status

Healthy Worker Effect

Loss to follow-up

Control selection: COVID-19 example

Restricting to individuals who went out and sought a clinical test may reduce selection bias, but...

People who seek testing when they are asymptomatic could differ from those who sought testing because they were symptomatic

These differences could be associated with other risk-related behaviors (masking, vaccination, etc...)

Selection Bias in Study of Pap Smear and Cervical Cancer (control selection)

Example:

Investigators are evaluating role of pap smears in the prevention of cervical cancer

Investigators identify 250 newly diagnosed cases of cervical cancer by reviewing medical records where cancer patients are treated

Select population controls among women who live in the same neighborhood as cases

Travel through neighborhoods on foot during weekday working hours to enroll 250 controls

Findings

No association between pap smears and cervical cancer OR = 1 •

HOWEVER..................

Women at home during day less likely to be employed

Less likely to have regular medical check-ups • Source population different from the cases

If controls were more appropriately matched True OR=0.44

Pap Smear and Cancer Interpretation

There is 56% reduced odds of having a pap smear among women with cervical cancer compared to controls (OR=0.44). Pap smear is protective for cervical cancer

Results with control selection bias were biased towards the null

Selection bias occurred because "being home during the day" was required for controls but not for cases

These criteria were related to exposure status (pap smear)

The control group did not accurately represent the exposure distribution in the source population that produced the cases

This is avoidable in the study design phase!

Non-response or refusal bias (agreement to participate is related to exposure and disease/outcome status (self-selection bias)

Example: Self-Selection Bias

Selection bias can be introduced into case-control studies with low response or participation rates if the likelihood of responding or participating is related to both the exposure and the outcome.

Ex: Diseased subjects who had the exposure had a participation rate of 80%, and the other three categories had participation rates of 60%.

Differential Surveillance, Referral or Diagnosis of Subjects

Example: Investigators conducted a case-control study to determine whether use of oral contraceptives increased the risk of venous thromboembolism (blood clot).

Case group consisted of women who had been admitted to the hospital for venous thromboembolism.

Controls were women of similar age who had been hospitalized for unrelated problems at the same hospitals.

The interviews indicated that 70% of the cases used oral contraceptives, but only 20% of the controls used them.

The odds ratio was 10.2

However...there had been earlier reports suggesting this association.

As a result, health care providers were vigilant re: their patients on oral contraceptives and were more likely to admit them to the hospital if they developed venous thrombosis or any signs or symptoms suspicious of thromboembolism.

The study had a tendency to over sample women who had both the exposure and the 3/6/23 outcome of interest

Selection bias due to over-sampling of women with the exposure and the outcome

Selection bias could have been minimized by more restrictive case selection criteria, such that only women who clearly required hospitalization would be enrolled in the case group

Self-Selection Bias in a Prospective Cohort Study

At time of enrollment nobody knows who will develop the outcome

All participants are part of the same cohort with common eligibility criteria

The entire cohort may be healthier than the general population. However, the comparison group is an internal one. It comes from the same healthy cohort

If healthy people are more likely to volunteer for a cohort study, one might question whether the results are applicable to the broad population

For the study population, the measure of association will not be biased (internal validity is ok)

External validity or generalizability is the bigger issue.

Self-Selection Bias in a Retrospective Cohort Study

Could occur in a retrospective cohort study, especially if individuals have to provide informed consent for participation

Since a retrospective cohort study starts after all cases of disease have occurred, subjects generally would know both their exposure and outcome status

Those with the most interest in participation could both have been exposed and have the disease

Another less common mechanism of selection bias in a retrospective cohort study might occur if retention or loss of records of study subjects (e.g., employment, medical) were related to both exposure and outcome status

Selection bias can occur if selection or choice of the exposed or unexposed subjects in a retrospective cohort study is somehow related to the outcome of interest

Example: Self-Selection Bias in a Retrospective Cohort Study

An investigation of an occupational exposure (e.g., an organic solvent) that occurred 15-20 years ago in factory.

Over the years there were suspicions that working with the solvent led to adverse health events, but no definitive data existed.

A retrospective cohort study was conducted using the employee health records.

If all records had been retained, the results might look like those shown below

However, suppose that many of the old records had been lost or discarded, but, given the suspicions about the effects of the solvent, the records of employees who had worked with the solvents and subsequently had health problems were more likely to be retained.

Consequently, record retention was 99% among workers who were exposed and developed health problems, but record retention was only 80% for all other workers.

Loss to Follow-up: Selection Bias

The differential loss to follow up resulted in a relative under-sampling of exposed subjects who developed the outcome Depending on which category is underreported as a result of differential loss to follow-up, either an underestimate or overestimate of effect (association) can occur

Ways to Minimize Loss to Follow-up

Maintain high follow up rates (>80%)

Enrolling motivated subjects

Enrolling subjects who are easy to track

Making questionnaires as easy to complete as possible

Maintaining the interest of participants and making them feel that the study is important

Provide incentives

Healthy Worker Effect?

A special type of selection bias that occurs in cohort studies of occupational exposures when the general population is used as the comparison group.

The general population consists of both healthy people and unhealthy people. Those who are not healthy are less likely to be employed, while the employed work force tends to have fewer sick people.

• People with severe illnesses would be most likely to be excluded from employment, but not from the general population

• Comparisons of mortality rates between an employed group and the general population can be biased

Example: Selective survival can also affect a cohort study

Two study approaches for study of natural history of HIV infection-effect of recruitment

(1) Recruit participants at the time they become HIV infected

(2) Recruit participants by means of a serologic survey to detect prevalent infections

People who progress to AIDS and die relatively soon after infection will be underrepresented in study 2.

The second study will underrepresent people who have an especially rapid course to death, since they will be underrepresented among prevalent cases.

Even if the second study is able to control for time since infection, there is a possibility that it will overestimate time to death compared to study 1.

Information Bias

Systematic differences in the measurement of exposure, outcome or other variables

'Flaw' in measuring exposure, covariate or outcome variables that results in different quality (accuracy) of information between comparison groups

Participant or investigator as source

Occurs during data collection and may be of several types

Measurement error that distorts the true association between an exposure and disease

Types of Information Bias

Interviewer Bias - an interviewer's knowledge may influence the structure of questions and the manner of presentation, which may influence responses

Recall Bias - study participants with a particular outcome or exposure may remember events more clearly or amplify their recollections

Key Features of Information Bias

Occurs after the subjects have entered study

It pertains to HOW the data are collected

Often results in incorrect classification of participants as:

Exposed or unexposed • Diseased or not diseased

Can create a bias either toward or away from the null

Can be avoided by careful design and conduct 3/8/23of study

Recall Bias

Recall Bias: those with a particular outcome or exposure may remember events more clearly or amplify their recollections

There is a differential level of accuracy in the information provided by compared groups

Occurs in case-control study if cases are more or less likely than controls to recall and report prior exposures (most likely scenario)

Occurs in a cohort study if exposed subjects are more or less likely than unexposed subjects to recall and report subsequent diseases

Not the same as 'poor recall' or 'limitations in recall'

Ways to Minimize Recall Bias

Ways to minimize or avoid its occurrence

Select a diseased control group

Assumes exposure is unlikely to cause a general increase in the risk of multiple outcomes

Design a structured questionnaire for complete and accurate exposure ascertainment

Questionnaire design is important

Self-administered (useful for socially sensitive topics.....examples? Issues with this approach?)

Biological measurements or medical record data

Is Recall Bias Present?

Extent in studies may be largely unknown

Ask subjects about knowledge of study hypothesis at end of an interview

Ex. Folic acid use and risk of neural tube defects

Have you heard there are any vitamins minerals or anything else that may cause or reduce the risk of birth defects? If yes, which one?

Those with knowledge of hypothesis might be more prone to recall bias

Interviewer Bias

Interviewer Bias: an interviewer's knowledge may influence the structure of questions and the manner of presentation, which may influence responses

Systematic difference in soliciting, recording, or interpreting information that occurs in studies using in-person or telephone interviews

Can occur in all epidemiologic studies

Case-control study: Interviewers are aware of the disease status of a subject and question cases and controls differently about their exposures

Cohort or experimental study: Interviewers are aware of the exposure or treatment status of a subject and question exposed and unexposed subjects differently about their diseases or other outcome

Ways to Minimize Interviewer Bias

Ways to minimize or avoid its occurrence

Mask interviewers to the participant disease or exposure status

Design standardized questionnaires

Closed-ended, easy to understand questions

Translation for non-English speakers

Avoid medical terms or difficult chemical names

Computerized Assisted Telephone Interview

Train Interviewers -Standardized question -Non-directive probing

Makes for a long interview!

Types of Information Bias: Differential Misclassification

Differential - the rate of misclassification differs in different study groups

Example: misclassification of exposure may occur more often in cases than controls

Types of Information Bias: Non-Differential Misclassification

Non-Differential - results from the degree of inaccuracy that characterizes how information is obtained from any study group - not related to exposure status or case/control status.

Example: problem with data collection methods.

Misclassification visuals

Non-differential misclassification bias towards null

• Non-differential misclassification blurs the distinction between groups.

  So the observed association gets weaker.

Differential misclassification could go in either direction

Information Bias

Error arising from systematic differences in the way info on exposure and disease is obtained from study groups

Participants are incorrectly classified

Occurs after subjects have entered the study

Retro and prospective studies

Recall and interviewer bias à misclassification

Potential for Selection Bias in Evaluating Screening Programs

4 types :

1) Referral Bias (volunteer bias, self selection bias):

Who was screened vs. who was not

Volunteers may be healthier than the general population; more likely to comply with medical recommendations. Different SES.

• or the opposite! We don't know

Potential for Selection Bias in Evaluating Screening Programs

4 types :

2) Length Bias (length-biased sampling, prognostic selection)

Occurs when screening tests disproportionately identify diseases with long preclinical detectable phases (the time disease is present but asymptomatic)

• Aggressive, fast-growing diseases have short asymptomatic windows and are more likely to show symptoms faster. Slow-growing diseases remain in the “detectable” window longer, make them move likely to be found during screening.

  • Screened group appears to have better survival rates even if the screening itself has no actual medical benefit.

Ex. the selective identification of more slowly developing cancers (less aggressive disease)

Potential for Selection Bias in Evaluating Screening Programs

4 types :

Lead time bias

The perception that the screen-detected cases has a longer survival because disease was identified earlier

• IF early detections is truly associated with improved survival, survival in the screened group should be greater than survival in the control groups plus the lead time.

Lead time bias occurs when early detection makes survival seem longer, even if death occurs at the same time. Length time bias happens when screening disproportionately detects slow-progressing, less fatal cases, overestimating overall survival rates

Potential for Selection Bias in Evaluating Screening Programs

4 types:

Overdiagnosis Bias

Type of misclassification bias

• Inflate estimate of survival

• Wrong conclusion about screening improving survival → screening can half false + → follow them → false inflated survival rate

Two categories of error: Random and Systematic

Random Error

Fluctuations around a true value of a parameter

Due to poor precision

Due to sampling error

Variability in measurement

Two categories of error: Random and Systematic

Systematic Error

Non-random

Measurement Biases